Everyday AI

Ep 836: Updated GPT-5.6, A new Cheap Meta Model, Qwen 3.8 released and impressive and 7 more AI updates you can use today

Everyday AI — Fri, 07 Aug 2026 15:39:14 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: OpenAI, Meta, and Alibaba all shipped major AI updates this week. We cover the features and models you can start using today. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Anthropic is teaming up with Millennium, Perplexity launched Computer for Builders, and OpenAI's AI device with Jony Ive is taking shape. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: ChatGPT is getting GPT-5.6 upgrades and unlimited free chats, ByteDance is training a frontier AI model, and Kimi K3 reportedly tried to access the internet during a security test. And more. Read on for Byte Sized News.

💪 Leverage AI: OpenAI just made one of its best AI models free for everyone. We break down that update and the other AI tools worth trying today. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: OpenAI: Our agents created secret message boards, Meta's agents (also) go rogue, Anthropic working on AI chips and more. Check it here!

Ep 836: Updated GPT-5.6, A new Cheap Meta Model, Qwen 3.8 released and impressive and 7 more AI updates you can use today

Free AI users rejoice: there’s finally unlimited, free frontier AI. 🤯

(That doesn’t require like a $50k machine to run.)

That’s because OpenAI just updated its GPT-5.6 model and with the update, it made GPT-5.6 Luna available to free users without limits.

And that wasn't the only big win this week for free and cheap AI.

From open weight models and an Open Agent Plugin standard to Meta's new pricing structure and coding harness, we saw a lot of new AI features this week.

Here's the 7 you should be using today.

Also on the pod today:

• Unlimited GPT-5.6 for free users 🤯
• Meta’s contributor tier: 95% off 💸
• Adobe Creative Cloud unified plugin 🎨

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – BrowserOS neo is The Missing Browser for Claude, Cowork & Codex, Progress AI Observability is AI Observability for Production Agents, Rindler signs into the sites your team already uses and does the busywork of a room full of operators.

Anthropic and Millennium — Millennium is teaming up with Anthropic to build an AI-powered risk analyst that helps

Google WeatherNext AI — Google’s WeatherNext AI just gave cyclone forecasts a full day’s lead time boost, and now the model is open source.

Google Maps AI — Google Maps just got way smarter, letting you order food, get real-time transit updates, and even remember your past plans.

Perplexity Computer for Builders — Solo founders are shipping faster with Perplexity Computer for Builders, a toolkit that connects code, payments, and reports in one place.

Suno AI Music — Suno is doubling down on building AI music tools that support real artists, not copycats.

Gemini Robotics 2 — Google DeepMind gave Apollo 2 the spotlight to share what it’s really like running on Gemini Robotics 2.

Fable 5 Upgrade — Fable 5 just got way better at handling biology questions, cutting false positives by 85 percent.

ElevenLabs Dubbing v2 — ElevenLabs just launched Dubbing v2, letting you dub audio with real emotion and accurate accents in 90+ languages

Jony Ives AI Gadget — OpenAI is teaming up with Jony Ive on a doughnut-shaped, battery-powered smart speaker that could cost over $300.

OpenAI Stats — OpenAI just dropped country-by-country data showing that ChatGPT is getting more global, more practical, and more popular with older users.

1. ChatGPT Rolls Out GPT‑5.6 Updates, Unlimited Free Text Chats 😋

OpenAI says Plus and Pro users can access the updated GPT‑5.6 Sol today, with a new slider to control how much time ChatGPT spends thinking through a response.

Free and Go users will receive GPT‑5.6 Luna this week, followed next week by unlimited text chats and a Think button for tougher questions.

2. ByteDance Trains AI Model Aiming Near Anthropic’s Mythos Scale ⚖️

ByteDance is reportedly training a major new AI system that could approach the scale of Anthropic’s frontier Mythos model, according to an exclusive Financial Times report published Friday.

If accurate, the effort signals that the TikTok owner is pushing harder into the top tier of global AI development, where only a handful of companies can afford the computing power and talent required.

3. China’s Kimi K3 Reportedly Tried to Reach the Internet During a Security Test 👮

According to WIRED, Kimi K3, an open-weight Chinese AI model, reportedly attempted to access the internet while being tested, apparently to get around the task it was given.

The finding matters because it adds to growing scrutiny of AI systems that can use tools and take multi-step actions beyond a chat window.

4. OpenAI Unveils Agent Plugin Standard With Major Tech Partners 🔌

OpenAI has announced Agent Plugins, a new open format meant to let developers package agent skills and MCP server settings once for use across compatible AI clients.

The effort was developed with companies including Amazon, Cursor, Microsoft, and Vercel, while GitHub helped refine the specification, according to posts cited in the discussion.

5. Fable 5 Slashes Biology Fallbacks 85% 🧑‍⚕️

Anthropic says it has updated Fable 5’s biology safeguards, reducing cases where biology questions are diverted to its less capable Opus 5 model by about 85%.

The change is aimed at allowing more everyday health, clinical, and educational support while still restricting requests involving dual-use areas such as virology, toxicology, molecular design, and drug development.

6. Trump Says Congress Could Regulate AI “Out of Business” ❌

According to Reuters, President Trump said Congress is pushing AI rules that could stifle the industry, sharpening Washington’s fight over how quickly to impose federal safeguards.

The comments arrive as OpenAI and Anthropic have reported troubling security-test failures, adding urgency to concerns that advanced systems can create real cybersecurity risks.

The cost of AI just changed.

Big time.

OpenAI just handed its billion free users unlimited GPT-5.6 Luna, a model hanging right around Claude Sonnet 5, a tier that caps paid users elsewhere at about 20 prompts every five hours.

"Free plans are always a tier behind, right?"

Yeahhh, that rule just died this week.

And that was only the loudest drop in the wildest free and open AI week ever: Adobe's creative arsenal inside ChatGPT, frontier coding at 95% off, open models crashing the top five.

We broke em all down on today's episode of Everyday AI, no technical chops required, plus which ones y'all gotta get reps on immediately.

Let's get into it.

1. Adobe Stuffed 70+ Tools Into ChatGPT 🎨

Adobe just collapsed its entire creative empire into one single ChatGPT plugin that taps more than 70 tools across Photoshop, Firefly, Acrobat, Express, Premiere, Lightroom, Illustrator, InDesign, and Adobe Stock.

The unlock? You never have to guess which app does what again.

Describe the outcome, dump your files in, and it routes everything automatically.

Guest mode even works without an Adobe account, though signing in unlocks the heavier generative firepower plus Creative Cloud sync.

Try This

Dump a folder of product photos and a messy spreadsheet into ChatGPT, then ask for a polished photo set plus a one-page PDF.

If it nails the routing on the first pass, you just deleted three app tabs from your creative workflow for good.

2. Meta Priced Frontier Coding At 95% Off 💸

Meta released Muse Code, a terminal coding agent powered by its brand new Muse Spark 1.2 model and a million-token context window, with standard pricing at $1.25 per million tokens in and $4.25 out.

The catch? A new contributor tier drops that to 10 cents in and 20 cents out, roughly 95% off, if you let Meta TRAIN its future models on your data.

Enterprises should sprint the other direction. But for hobbyists starting from scratch, honestly, not a terrible trade.

The bigger story: Anthropic reportedly makes about 80% of its revenue on coding, and Meta just aimed its cheapest weapon straight at it.

Try This

Got a personal project with zero proprietary code in it? Install Muse Code on the contributor tier and let a background agent grind through your weekend backlog.

Just keep anything client-facing or remotely proprietary far, far away from that tier.

3. Alibaba Crashed The Top Five Model Club 🏆

Riding that same open wave, Alibaba dropped Qwen 3.8 Max, a 2.4 trillion parameter open weight model that reads text, images, and video.

Per Artificial Analysis, it sits in the same tier as GPT-5.6, Claude Fable 5, Claude Opus 5, and Kimi K3, meaning two of the world's top five models are now Chinese open weights.

The weights drop within days, though a fresh Reuters report says Alibaba plans to start charging major enterprise users a share of their revenue next week.

You won't be self-hosting this one without a spare data center in the basement. The real win is pricing pressure that keeps the big US labs honest.

Try This

There's nothing for most of us to install this week unless you have serious compute collecting dust.

Instead, forward this one to whoever negotiates your AI contracts, because frontier-level open models are your best leverage at renewal time.

4. Gemini Notebook Finally Went Full Agent 📓

The tool formerly known as NotebookLM got its glow up: it's now Gemini Notebook, and for paid users, it's agentic by default.

Every notebook gets a secure cloud computer that writes and runs real code against your sources, then hands back charts, spreadsheets, slide decks, and PDFs.

It runs on Gemini 3.5 with the Antigravity harness plus 100 built-in skills, and it can even query the web now, asking permission first so your grounding stays intact.

One catch: under the hood sits Gemini 3.5 Flash, scoring a 52 on Artificial Analysis, roughly what free ChatGPT users now get unlimited.

Access: Gemini Pro accounts, web only.

Try This

Load last quarter's reports into a notebook and ask for the charts plus a slide deck in one prompt.

If it saves your analyst a full afternoon, we did the complete walkthrough in episode 834.

5. Agent Plugins Became The Universal Standard 🔌

There's a new open, vendor-neutral standard for packaging AI agent extensions: one folder with a plugin.json file bundling your agent skills and MCP servers together.

Build a plugin once and it runs across ChatGPT, Codex, Cursor, GitHub Copilot, VS Code, and more, with Vercel, Amazon, Anysphere, GitHub, and Microsoft all on board.

That matters because loading these into every tool separately burns hours and quietly locks your whole AI stack into one vendor.

Who's missing? Anthropic.

Almost every other big name jumped on within hours, which makes the lone holdout look like a lock-in play.

Try This

Point your coding agent at agent-plugins.org and repackage your team's existing skills and MCP servers into one portable plugin.

If part of your company ever needs a different AI stack, future you says thanks.

6. Cloudflare Gave Away Its Company Brain 🏢

Speaking of open, Cloudflare open sourced Cloudflare OS, the internal AI workspace it built to actually run its own company, and thousands of its employees already use it daily, so this ain't demo software.

Every employee gets a browser-based agent workspace: governed access to internal systems, research tied to live company data, automated workflows, and shareable internal apps with zero developers needed.

It competes head-on with ChatGPT Work, Claude Cowork, and Gemini Enterprise, except it's free, open source, and runs in your own Cloudflare account with your own model and spend caps through Cloudflare AI Gateway.

Try This

Send the repo to your IT lead with one question: what would a two-week pilot actually cost us?

If the answer comes back as mostly staff time, you're looking at a small fraction of what per-seat enterprise AI pricing runs.

7. MiniMax Open Sourced The Best AI Video 🎬

MiniMax pushed the open wave into video, releasing the weights for H3, the first open model to ever top a major AI video ranking: number one in video editing on Artificial Analysis, number two in text to video.

It reads text, images, video, and audio as one context, then spits out four to 15 second clips at up to 2K with native stereo audio.

Seen suspiciously good AI video flooding your feed this week? This was the engine.

The gnarly part? The Hugging Face weights exclude the US, EU, UK, and South Korea from local use, so those regions are API only.

And video this good before a US election should worry everyone.

Try This

Run one 15-second product or explainer clip through the hosted API before your next campaign brief.

If the quality holds up, your video production budget just changed for good.

8. Free ChatGPT Became AI's Best Deal 🎁

OpenAI capped the week by giving its billion users unlimited GPT-5.6 Luna on the free plan, replacing the old GPT-5.5 and GPT-5.4 combo entirely.

Luna hangs right around Claude Sonnet 5 territory, and free users now get it with NO usage cap.

Paid users get a cleaner setup too: instant and thinking merged into one model, with a slider controlling how much reasoning goes into each answer.

OpenAI also claims 68% fewer factual errors on financial, medical, and legal prompts versus GPT-5.5 Instant.

The subtext, fam: between cheap open models and free frontier-adjacent chat, it has been a rough stretch for Anthropic ahead of a public listing expected in as soon as two months.

Try This

Paying for seats your team barely touches? Have your light users re-test free ChatGPT this week and count who actually needs the paid tier.

If half your seats can downgrade, that budget just moved to tools doing real work.

Ep 835: Inside Everyday AI: My 9 most Used AI Tools and Workflows

Everyday AI — Thu, 06 Aug 2026 15:24:33 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: From ChatGPT Work to Gemini Notebook, we break down the nine AI tools and workflows powering Everyday AI. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Qwn 3.8 overtakes Anthropic's Opus, Meta's Muse Spark 1.2 jumped into the top tier of AI models, Grok upgraded Voice Mode with connectors, and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: OpenAI: Our agents created secret message boards, Meta's agents (also) go rogue, Anthropic working on AI chips and more.. Read on for Byte Sized News.

💪 Leverage AI: Thousands of hours of AI use boiled down to our 9 most important AI workflows and tools we use. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Google Deepmind CEO is gone from post, Safety report finds 19 agent break outs, SpaceX and NVIDIA plan orbital AI compute. Check it here!

Ep 835: Inside Everyday AI: My 9 most Used AI Tools and Workflows

"How do you use AI every day?" 🤔

One of the most common questions I get. And TBH, it's been hard to give a straight answer.

Because by the time I've collected all the info, something changes. But over the last few months -- as work goes from the web to desktop -- my personal workflows have steadied and stayed consistent.

So I'm taking you inside Everyday AI and the 9 AI tools and workflows I use the most.

Hundreds of AI tools and modes a year.

Thousands of hours.

All boiled into one episode.

Also on the pod today:

• Codex dominates desktop AI workflows 💻
• Overnight agents crunch podcast stats ⏰
• Automated comms triage saves mornings 📬

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Wispr Flow Notetaker is Meeting notes that get the details right, NgrokAI is One gateway for every model, Cloudflare Wallets is the programmable wallet for the agentic Internet

Meta Muse Spark 1.2 — Meta’s Muse Spark 1.2 just jumped to the top tier for agentic knowledge work, matching Grok 4.5 and GPT-5.5 at a bargain price.

Grok Voice Mode Connectors — Grok just leveled up voice mode with connectors, so you can talk to your emails and apps.

Codex and Edge — ChatGPT’s Computer Use feature now works in Microsoft Edge, with a new settings menu for more browsers.

Qwen 3.8 — Alibaba’s Qwen 3.8 just took the top spot on Artificial Analysis’s Agentic Index from Anthropic’s Opus 5.

OpenAI Ads — OpenAI is rolling out ads on ChatGPT in Brazil and Mexico, with new features like conversion optimization and carousel formats.

ByteDance SeedRealTime — ByteDance just dropped SeedRealtime, a next-gen AI that can watch, listen, and chat in real time on Doubao.

Reddit AI — Reddit’s rolling out smarter moderation tools and tighter security, plus big changes for developers and legacy systems.

1. Meta Model Breaches Company Systems 👾

Meta said Wednesday that a testing misconfiguration by partner Irregular accidentally gave one of its models internet access, allowing it to exploit a flaw in a third-party service.

According to The Information, the model was Muse Spark 1.1 and it altered an unidentified company’s internal systems, though Irregular says the episode was an already-known evaluation-environment issue, not a sandbox escape.

2. Anthropic Starts Hiring to Build Custom AI Chips ⚙️

Anthropic has confirmed it is assembling a custom silicon team to design chips tailored to running Claude faster and with less power.

The move, first reported by Business Insider, signals that the AI company wants more control over the costly computing capacity behind its rapidly growing service.

3. Meta Unveils Muse Code 🧑‍💻

Meta has launched a preview of Muse Code, its first coding agent, as Alexandr Wang’s AI division pushes the company further into paid developer tools.

According to Bloomberg, the agent can plan software changes, write code, and validate results, pairing with Meta’s Muse Spark 1.2 model to compete with products such as Anthropic’s Claude and OpenAI’s Codex.

4. DeepSeek Signals Significant API Price Increase 📈

DeepSeek has added a notice to its API platform warning that prices will rise “significantly” soon, according to screenshots shared by TestingCatalog and AiBattle.

The move could weaken one of DeepSeek’s biggest draws: unusually low-cost access to capable models for developers and small AI products.

5. OpenAI Says Rogue Agents Used a Message Board While Hacking Companies 🛡️

At Black Hat this week, OpenAI disclosed new details about an incident in which its AI agents allegedly planned hacking activity through a message board without the company noticing.

According to WIRED, the agents breached several outside companies during what appears to have begun as a testing effort, raising fresh questions about how closely powerful AI systems are monitored once they gain access to tools and networks.

The most productive worker here clocks in at 11 PM and logs off at 5 AM. Not a person, a squad of AI agents running third shift inside OpenAI's Codex.

Third shift agents?

Yep. They crunch stats, triage inboxes, and tackle the day's toughest problems while everyone sleeps, then leave a finished briefing by sunrise.

You know.... "Codex" has "code" in it, sounds like developer stuff.

Wayyy wrong, fam.

This setup has run nightly since February, after thousands of hours on every $200 plan out there. The models are prolly a coin flip now, but the harness race ain't close.

What's on today's show? The full nine-workflow playbook, from the overnight briefing agent to the one-sentence skill habit to the voice trick that runs a computer mid-walk.

Zero code required, promise.

1. Overnight Codex agents end morning busywork 🔥

Every night at 11 PM, scheduled agents clock in. They pull podcast stats from Buzzsprout, check newsletter A/B tests inside Beehiiv, and triage four inboxes plus LinkedIn with replies drafted.

By 5 AM? One clean morning briefing, sitting there waiting.

The trick that makes all of this work is a feature called browser control. When a tool has no clean integration, the agent opens Chrome and clicks around like a patient employee, so NOTHING on your stack is off limits.

That job used to burn an hour of pulling, exporting, and squinting at dashboards. Now it's 30 seconds with coffee.

The best part? One evening of setup, and the payoff repeats every morning after.

Try This

Open Codex tonight, or the ChatGPT desktop app since they're the same thing, and describe your most dreaded recurring report: where the numbers live, which logins it needs, and what the finished briefing should include. Schedule it for 11 PM.

The first run will be messy, so correct it once and save the instructions. By Friday, the report builds itself.

2. AI skills turn repetition into owned IP ⚡

Forget counting tokens or lines of code. The real measure of an AI native operator is how many saved skills fire without a human touching anything.

What's a skill? Any multi-step task you've done twice, saved as repeatable steps, and chained into a plugin an agent reruns on a schedule.

The magic words are stupidly simple: after finishing any task, type turn this into a skill, and the agent saves the whole process.

The current count around here: 124 skills, used a combined 6,179 times. One of them scans every single thread daily and flags repeat work that should become the next skill.

Skills building skills. Wild, right?

Try This

Start tomorrow morning with one simple rule: anything you do by hand twice gets the turn this into a skill treatment on the spot. It takes under a minute and compounds every single day.

After a week, stack your best three skills into one plugin and schedule it to run overnight while you sleep. You gotta stop being a button pusher.

3. ChatGPT Voice runs your whole computer 🚀

ChatGPT Voice through Remote might be the most slept-on release of the year. Connect the mobile app to your desktop once, open Remote, tap the blue button, and your phone now runs the ENTIRE machine.

What does that actually unlock? Taking a walk while an agent opens the PowerPoint in your downloads folder, swaps in a fresh Canva image, reads back page three, then updates HubSpot and pings the team in Slack.

It's full duplex, so you can talk over each other while a second model keeps working in the background.

One gotcha: regular voice mode in the ChatGPT app can't touch your connectors, but voice through Remote can.

Try This

Connect the ChatGPT mobile app to your desktop, then open Remote and tap the blue button. On your next 20-minute walk, hand it one real task out loud, something like open the file in the downloads folder and tell me what changed.

Start small to build trust, then hand it more every single day. Y'all, it genuinely feels like cloning yourself.

Ep 834: Gemini Notebook: 7 New Updates and What They Unlock

Everyday AI — Wed, 05 Aug 2026 17:36:42 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Google just gave Gemini Notebook its biggest upgrade yet. We break down the new features and what they unlock for your workflow. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Gemini Notebook just got a major upgrade, FLUX 3 Video is now available to everyone, and Qwen-Image 3.0 has officially launched. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Google Deepmind CEO is gone from post, Safety report finds 19 agent break outs, SpaceX and NVIDIA plan orbital AI compute. Read on for Byte Sized News.

💪 Leverage AI: NotebookLM is gone. Gemini Notebook is in its place. Here’s the 7 new updates you should know. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: OpenAI challenged Apple's trade-secret claims, the White House is launching AI cyberattack tests, and Gemini Notebook reached all Pro users. And more. Check it here!

Ep 834: Gemini Notebook: 7 New Updates and What They Unlock

Not only does NotebookLM have a new name, it's got a new game.

Gemini Notebook is agentic by default, can think and reason, and can output files now in just about any format.

On this week's AI at Work on Wednesday, we show you the 7 New Updates in the new Gemini Notebook, how they work, and how you should use them.

Also on the pod today:

• NotebookLM renamed to Gemini 📒
• Collections for notebook organization 🗂️
• Automatic Google Drive syncing 🔄

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Hey Noah is the AI that schedules better than your EA, Atlaso is One shared memory layer for every AI tool you use, SpeakoFlow is Free, open-source voice-to-text.

Gemini Notebook Upgrades — Google’s NotebookLM just got a big upgrade for Pro and Ultra users. Now you can generate files and do research straight from chat

FLUX 3 Video — FLUX 3 Video just launched for everyone, letting you create HD videos with audio straight from text or images.

Autodesk Flow Studio — Autodesk Flow Studio just launched a 3D editor that lets you direct scenes in true 3D, then polish the final look with AI.

Qwen-Image-3.0 Released — Qwen-Image-3.0 just launched, ranking top among Chinese and mainstream image models.

Spotify AI — Spotify is rolling out a paid AI tool that lets fans create covers and remixes with artists' permission, aiming to give both credit and cash to musicians.

AI in the Classroom — Three new ChatGPT education plugins aim to bring course materials and school tools into one place.

Amazon and Perplexity — A federal court just sided with Perplexity, letting its AI agents shop on Amazon again.

GPT-Rosalind Codex Updates — OpenAI is testing a public download page for its life sciences-focused GPT-Rosalind Codex, hinting at wider access beyond select research partners.

Fireworks AI — Fireworks AI models can now be deployed in Azure with Microsoft Foundry, making it easier for startups to launch and scale open model inference without managing infrastructure.

National AI — The NSF is rolling out AI hubs so more colleges can tap into top-tier computing, with NVIDIA backing the move.

1. White House Exempts Open-Weight AI Models From Voluntary Safety Tests 🏋️

According to Reuters, the Trump administration told major AI companies Tuesday that publicly available open-weight models will not face its planned voluntary cybersecurity testing.

The move separates systems such as Meta’s Llama and Nvidia’s Nemotron from more tightly controlled models made by OpenAI, Google, and Anthropic, even as concerns grow that capable AI can help carry out cyberattacks.

2. SpaceX, NVIDIA Plan Orbital AI Compute 🛰️

SpaceX and NVIDIA announced that the Starmind AI1 satellite will carry an NVIDIA Vera Rubin NVL72-powered compute payload.

The project aims to use solar energy in space to support AI computing on Earth, moving power and cooling constraints far above the atmosphere.

3. UK AI Safety Tests Flag Unauthorized Agent Actions 🚩

Reuters reports that Britain’s AI Security Institute found 19 unauthorized actions across tests of OpenAI and Anthropic agents, including one case involving malicious code and fake online identities aimed at securing human approval.

Anthropic’s agent accounted for 17 incidents, while OpenAI reported two cases in which its agent used the internet beyond the test prompt’s rules.

4. Google DeepMind CEO Demis Hassabis shifts to chairman role as Jeff Dean exits 🫡

Hassabis is handing daily control of Google DeepMind to CTO Koray Kavukcuoglu while becoming Alphabet’s chief scientist and continuing to lead Isomorphic Labs.

The shake-up also sends longtime Google scientist Jeff Dean and several prominent AI researchers out to form Discovery Loop, a new company backed by Google. The changes arrive as Google faces pressure to move Gemini forward faster and retain key talent in an increasingly fierce race with OpenAI and Anthropic.

5. Hark Unveils Web-Using “Handoff” Agent, Claims Benchmark Lead 🧑‍💻

Hark announced Handoff, a computer-use agent it says can navigate websites and complete online tasks through its own virtual browser. The company, founded by Brett Adcock, who also founded robotics company Figure AI, claims the system set a new high score on the OM2W web-use benchmark, ahead of comparable models from Anthropic, OpenAI, and Google, while costing less to run.

Hark says Handoff is designed to handle browser-based work from start to finish, marking another push to turn AI chat tools into agents that can actually operate software.

6. Cloudflare Opens Its In-House AI Workspace as Cloudflare OS ☁️

Cloudflare has launched Cloudflare OS, an open-source browser workspace that lets employees build and share AI-powered tools using their company’s own data and systems.

According to reports, the platform is designed to move AI work beyond centralized developer teams, allowing staff to research, create live-data documents, and automate routine workflows without coding. Companies run it in their own Cloudflare accounts, keeping control of the apps, connections, and workplace context they create.

NotebookLM is dead.

(We’ll miss you, dear friend.)

What replaced it might be the easiest AI win your company grabs all year.

Google graduated its cult-favorite research tool into Gemini Notebook, and has rolled out seven upgrades that turn it agentic by default.

Translation: it stopped waiting for orders and started doing real work. The goodies locked behind Google's priciest plan just hit every paid account, prolly yours too.

Don't worry: none of it needs code or engineers.

That's what we tackle on today's episode of Everyday AI: what the rebrand changed, the one-prompt move that builds five deliverables mid-meeting, and the grounding shift that could burn careless teams.

It's Wednesday, y'all, so let's put AI to work.

1. NotebookLM dies, Gemini Notebook goes full agent 🚀

Everybody assumed this tool was next for the Google graveyard, but instead it got promoted into the flagship Gemini family. That promotion matters way more than the new logo: you can finally build serious workflows without fearing a random shutdown email.

Under the hood, the changes run a whole lot deeper than the name suggests. Old NotebookLM ran a non-reasoning model with a fixed output menu, while the new one runs Gemini 3.5 reasoning, Google's Antigravity harness, and a secure cloud computer in every notebook.

So why should busy leaders care? Because your notebook writes and executes code for research, calculations, and data analysis, grounded in YOUR sources, not the open internet.

Try This

Drop something real into a fresh notebook this week, meeting notes or a market report, and ask one question that needs calculation, not just a summary.

Then open the thoughts panel and watch it reason, run code, and cite your sources in real time, a free masterclass in how agents think.

Just 10 minutes there beats a month of headlines.

2. One prompt now builds five executive deliverables ⚡

This is where agentic stops being a buzzword and starts printing time back into your week. We threw it a beefy live test: analyze three AI budget strategies for a 5,000 person enterprise, run the cost math, pick a winner, and then build five deliverables in one run.

What came back? A custom audio overview, a mind map, an editable Excel workbook with working formulas, a Nano Banana infographic, and a two-page executive brief, every claim traceable to the source notes.

The whole package landed in about six minutes, versus an afternoon of assembly across five tools, and that kind of math changes how fast your whole team ships decisions.

Try This

Grab one real decision off your desk, feed the background docs into a notebook, and ask for the whole package in one prompt: analysis, spreadsheet, exec brief, and infographic.

Then open the Excel workbook it hands back, because yes fam, the formulas actually work.

Once your most skeptical teammate watches a finished package appear mid-meeting, the buy-in conversation ends itself.

3. The 100% grounding promise just got an asterisk 🔥

Quick context: executives trusted NotebookLM because it answered from your sources ONLY, crushing hallucinations in a way OpenAI, Anthropic, and Microsoft never matched, and Gemini Notebook keeps that grounding by default.

Now here's the gnarly part, though. When your sources come up short, it can offer to hop online and pull outside info into your notebook, and once you say yes, outside data starts mixing with yours.

The guardrail is real here, at least for now, since it always asks your permission before searching anything.

But if your team assumes every answer still comes purely from your documents, that assumption became conditional, and leaders gotta know which mode they're in.

Try This

Stress test this yourself by asking a notebook something your sources definitely can't answer, then watch it flag the gap and request permission first before touching the web.

After that, set one line of team governance that keeps the hallucination protection you signed up for: grounded by default, web only with approval, chain of thought checked when stakes are high.

Ep 833: RSI Explained: When AI Starts Improving Itself and What It Means

Everyday AI — Tue, 04 Aug 2026 17:31:19 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Recursive self-improvement is quickly becoming one of AI's biggest stories. We break down what it is, why it matters, and what's changing because of it. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: The EU is requiring AI deepfake labels, Gemini Spark is getting Chrome integration, and ChatGPT is getting real-time voice. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: OpenAI challenged Apple's trade-secret claims, the White House is launching AI cyberattack tests, and Gemini Notebook reached all Pro users. And more. Read on for Byte Sized News.

💪 Leverage AI: OpenAI just used AI to make AI cheaper. We break down what recursive self-improvement means for your business. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Report: Dario worried Anthropic hires care too much about money, OpenAI teases Astra: new model family, Qwen 3.8 and Deepseek V4-Flash impress and more. Check it here!

Ep 833: RSI Explained: When AI Starts Improving Itself and What It Means

Intelligence too cheap to meter -- could that actually be coming? 🤑

Maybe, and you mighta missed one of the biggest signs.

OpenAI said that its own model improved itself after release so well, that it was cutting prices to one model by 80%.

What's that mean?

Well, by definition, it's a broad example of Recursive Self Improvement, or when a model starts building better versions of itself.

And over the past 8 weeks, RSI has gone from a science fiction future to an actual, near-term reality.

So what happened and what changed? And what does it mean for your company?

We break it all down.

Also on the pod today:

• OpenAI drops AI prices 80% 💸
• AI models now upgrade themselves 🤖
• Recursive Self-Improvement (RSI) explained 🔁

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Airtop sets up campaigns, audits spend, and optimizes, Claudemon has Wild Pokemon appear while you work in Claude Code, PassiveShorts is Faceless short videos on Auto-Pilot.

EU AI Deepfakes — The EU just rolled out new rules forcing companies to label AI-generated content and deepfakes.

Gemini Spark and Chrome — Gemini Spark just got smarter with Chrome web browsing superpowers, now helping with online tasks like booking flights.

Gemini Enterprise Upgrades — Google is cooking up Plugins and a Notifications hub for Gemini Enterprise, hinting at smarter, reusable workflows.

MiniMax-H3 Arena — MiniMax-H3 just shot to #1 as the top open model in both Text-to-Video and Image-to-Video.

Grok on IOS — SpaceXAI is giving each Grok voice its own color and style, hinting at custom voice agents on iOS soon.

NVIDIA AI Push — Nvidia’s push for open AI models isn’t just about principle, it’s a calculated move to sell more chips as AI spreads everywhere.

SSI — SSI is rumored to be releasing its first model this month.

ChatGPT Realtime Systems — GPT-Live ditches turn-based voice AI for true real-time conversations, letting it listen and talk at once.

Cybersecurity Transparency — Over 120 orgs in the Open Secure AI Alliance just dropped new guidelines and open source tools to level up agentic AI cybersecurity

F1 and AWS — F1 cut data onboarding from 8 weeks to just hours by using AWS AI agents to automate and unify fan data management.

S&P 500 Record — The S&P 500 just broke another record, thanks to strong earnings from AI-focused companies and optimism about a potential Mideast deal.

1. OpenAI Publicly Challenges Apple’s Trade-Secret Claims 🍎

OpenAI has pushed back publicly against Apple’s recent lawsuit, calling its claims against former Apple executives Chang Liu and Tang Tan inaccurate and unnecessary.

In a new blog post, OpenAI released selected messages and emails to dispute Apple’s allegation that the pair took confidential information to support OpenAI’s hardware efforts.

2. White House meets with AI leaders for new AI framework 👾

The White House will meet Tuesday with major AI companies to discuss a newly completed voluntary framework for testing the cyberattack potential of the most advanced models, CNBC reports.

The program, ordered by President Trump in June, would let participating developers give the government up to 30 days of early access to qualifying systems before sharing them more broadly with trusted partners.

3. Gemini Notebook’s upgraded experience reaches all Pro users 😋

Gemini Notebook says its more powerful experience has now fully rolled out to every Pro subscriber, expanding support for files such as PDFs, images, documents, and spreadsheets.

The update also adds concurrent artifact generation and sharper chat-based analysis, positioning the tool as a more capable workspace for turning source material into usable outputs.

4. Nvidia unveils Alpamayo 2 Super for autonomous vehicles 🚗

Nvidia CEO Jensen Huang announced the open reasoning model on X, positioning it as a key step toward AI systems that can handle the messy, split-second decisions of real-world driving.

Alpamayo 2 Super is designed to interpret road scenes, reason through complex conditions, and support robotaxis, trucks, delivery fleets, and other mobile robots.

5. Palantir’s Q2 Surge Puts “AI Sovereignty” at Center Stage 📈

Palantir reported 93% year-over-year revenue growth in the second quarter, lifting its full-year outlook to roughly $8.15 billion as U.S. commercial sales jumped 149%.

CEO Alex Karp said the momentum reflects companies seeking tighter control of their data, intellectual property, and AI workflows rather than handing them to major model providers.

6. Anthropic Reportedly Signs $10 Billion Volta Infra Computing Deal ⚡

According to Bloomberg News, Anthropic has reportedly committed $10 billion to secure computing capacity from infrastructure startup Volta Infra, a major new bet on the hardware needed to run and train advanced AI systems.

The agreement highlights how the race to build more capable AI models is increasingly constrained by access to data centers, chips, and power, not just software talent.

The business world spent the last few months bracing for AI to get more expensive. Then AI cut its own prices.

Wait, what?

GPT-5.6 Sol improved its own smaller sibling Luna after release, and OpenAI passed the savings on with Thursday's 80% price cut.

That's a hint of recursive self-improvement, or RSI, which just means AI now builds better, cheaper AI while humans kinda just supervise.

You know the reaction: "Cool, another acronym for the AI alphabet soup."

Yeahhh, except when full RSI lands, the implications of acronym kinda dictate what your company can afford next quarter.

The sudden surgence of RSI might be the biggest shift in AI economics since ChatGPT dropped.

The gnarly part? The people building self-improving AI are the same ones telling Washington lawmakers that we might need to pump the breaks eventually.

On today's Everyday AI, we unpack why prices are falling when everyone swore they'd climb, the eight weeks that rewrote the timeline, and why the builders want a brake pedal, plus the moves leaders gotta make now.

Let's dive in shorties.

1. OpenAI's model made itself 80% cheaper 🔥

For months, the smart money said 2026 was the year to tighten token budgets. Powerful agents, hungry models, climbing bills.

Then Sol adapted the post-training setup for its smaller sibling Luna, and OpenAI slashed Luna's price by 80% and Terra's by 20%.

So what does that mean for your budget?

Luna now hangs with Claude Sonnet 5 on independent benchmarks per Artificial Analysis, at a fraction of the cost per completed task. Frontier-level smarts just hit the bargain bin.

Labs that properly invested in compute can keep squeezing costs out of their own models, so plan for AI prices to FALL and treat budgets built on climbing costs as stale.

Try This

Pull a list of every task running on your priciest model this week. Route the busy work, the summaries, drafts, and simple lookups, to a cheap small model like Luna, and save the flagship for judgment and strategy.

That's exactly how the labs run their own shops, so rerun the math monthly, because the price floor keeps moving under your feet.

2. Eight weeks that rewrote the AI timeline ⚡

RSI went from research jargon to the AI industry's main character in eight weeks.

Anthropic published research on AI building itself, and Google researchers mapped RSI as a main route to superintelligence, while OpenAI built an internal benchmark scoring how well models improve THEMSELVES.

Then it got wild, fam. Sam Altman declared we're in the singularity, a startup named Recursive signed a nine-figure AWS deal to automate AI research, and a Google DeepMind exec called those data-center bills an RSI bet.

Why do timelines matter?

Full RSI looked like a 2029 story, but now it's tracking more like 2027, with METR showing AI task length doubling every four months.

Translation: your competitors' capabilities won't shift yearly anymore. They'll shift quarterly.

Try This

Brief your leadership team on RSI this week, in words everyone gets, before the headlines force the conversation. Then run one scenario: what changes if AI costs drop 50% next quarter while capabilities jump?

Capture the two or three projects that suddenly pencil out. Teams that pre-game the price drops move the day they land, while everyone else schedules another meeting.

3. The people building RSI want brakes 🚀

Here's the plot twist nobody had on their bingo card: the same companies pouring hundreds of billions into RSI compute employ the folks asking to slow it down.

More than 1,300 frontier-lab employees, including top execs at the biggest labs, signed a letter called Pacing the Frontier. It asks the US government for tools to pace automated AI development later, not a pause today.

(And nope, China prolly ain't signing up.)

The worry underneath? A self-improving AI gets better at whatever its scoreboard rewards, even when that scoreboard is wrong.

Recent agent containment breaks were harmless. Once RSI hits full speed, they might not be.

Humans still own the judgment role, and your org chart should say so.

Try This

Name a fallback AI provider this month, even if you never switch, because RSI winners cut prices first and you want freedom to chase the value. Then kill the annual AI roadmap and plan in monthly cycles.

Finally, assign one human owner who signs off on what your AI produces. A self-improving model still needs a trustworthy scorekeeper, and that scorekeeper is you.

Ep 832: OpenAI’s new Astra model, more AI agents escape sandboxes, AI leaders call for AI pacing and more.

Everyday AI — Mon, 03 Aug 2026 18:13:20 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: OpenAI slashed GPT-5.6 prices, AI agents escaped more sandboxes, and industry leaders are calling to pace AI development. We cover last week's biggest stories. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Gemini's desktop app is getting major upgrades, Google pulled its Earth AI tool, and Alibaba climbed to #2 in the Text Arena. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Report: Dario worried Anthropic hires care too much about money, OpenAI teases Astra: new model family, Qwen 3.8 and Deepseek V4-Flash impress and more. Read on for Byte Sized News.

💪 Leverage AI: OpenAI slashed GPT-5.6 prices, AI agents escaped more sandboxes, and AI leaders called to pace development. We break down last week's biggest AI stories. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Anthropic found Claude reached real systems during testing, OpenAI cut GPT-5.6 prices, and the EU is investing €10 billion in AI gigafactories. And more. Check it here!

Ep 832: OpenAI’s new Astra model, more AI agents escape sandboxes, AI leaders call for AI pacing and more.

OpenAI has a new model coming soon called Astra.

Was it a leak?

A reddit post?

Some backdoor update?

Nope, OpenAI made some crazy discoveries and math then told the world that their next model family Astra did the heavy lifting.

(And you thought you could just click ‘Sol’ and your strategy was set for Q3?)

Aside from news on what’s next from OpenAI, this week saw multiple new agent outbreaks, AI competitors banning together to pace AI, Amazon doing a 180 on its AI strategy and a lot more.

Don’t get left behind. We’ll keep you ahead.

Also on the pod today:

• Anthropic, OpenAI agents escaped sandboxes 🤖
• OpenAI’s 80% model price cut 💸
• Claude models breached test environments 🛡️

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Capptivo is a Screen Studio and Cursorful alternative, Termexo is Claude Code in one terminal, Uniweb Adds Payments to Your Product in One Prompt Safely

Gemini Desktop Updates — Google Gemini's desktop app is getting image and video generation tabs, camera capture, and easier connector management

Google Earth AI Pulled — Google Earth’s new AI image tool was pulled almost immediately after launch, since users quickly started generating fake disaster images.

Grok Imagine Updates — Grok Imagine just dropped voice consistency, multi-references, and native 1080p for SuperGrok Plus and Heavy users.

CuspAI Funding — AI is now helping discover new materials for future computer chips, and CuspAI just raised $450 million with backing from Jeff Bezos, Nvidia, and AMD.

Alibaba Arena — Alibaba’s Qwen just jumped to #2 in the Text Arena rankings, edging out GPT-5.6.

OpenAI Math — OpenAI's new model cracked ten big math and computer science problems, including a disproof of the Erdős unit-distance conjecture.

AI Shakeup — Nobel Prize winner John Jumper is leaving Google for Anthropic, shaking up the AI world.

AI Studio Artifacts — Google AI Studio is finally getting an Artifacts folder so your files stick around, not just your session.

Microsoft Tests MAI Realtime — Microsoft is quietly testing MAI Realtime, its first native real-time voice model that listens and speaks at the same time.

1. Alibaba Unveils Qwen3.8-Max 🤯

Alibaba has released Qwen3.8-Max, its largest AI model yet, in a fresh challenge to leading US developers.

According to Bloomberg, the company says the 2.4-trillion-parameter system matches or sometimes exceeds Anthropic’s Fable 5 on selected benchmarks and outperforms Moonshot’s recently announced Kimi K3 in several tests.

2. OpenAI Hints at “Astra,” Its Next Major AI Model, in Math Research Post ➗

OpenAI quietly revealed that an internal version of a forthcoming model called Astra produced ten new mathematics and theoretical computer science results, according to Gizmodo.

The mention suggests Astra may be the next entry after GPT-5.6 Sol, though OpenAI has not confirmed its public name, timing, or whether it will retain the GPT label.

3. ChatGPT Captures 88% of Identifiable House AI Spending 🤑

According to CNBC’s review of House records through March 31, OpenAI’s ChatGPT accounted for about $100,580 of the $113,740 spent on named AI tools, far outpacing Anthropic’s Claude.

The figures offer an early snapshot of how Congress is adopting the same technology it will soon be asked to regulate, with staff using AI for briefs, research, constituent work, and hearing preparation. The real total is likely higher because free tools, bundled products like Microsoft Copilot, reimbursements, and Senate use are mostly hidden from public records.

4. DeepSeek’s V4-Flash Undercuts Global AI Rivals on Cost 💸

DeepSeek has released V4-Flash, a new flagship model that Reuters reports is the cheapest major AI system to run in Artificial Analysis testing, costing roughly 3 cents per test.

The model is estimated to be about 105 times cheaper than Anthropic’s Claude Fable 5, giving the Beijing startup a sharp price advantage as Chinese AI companies compete for global business users.

5. Amazon Hits $3 Trillion Market Value on AI Stock Surge 📈

Amazon briefly crossed the $3 trillion market-cap mark for the first time Today as renewed enthusiasm around artificial intelligence pushed major technology shares higher.

According to Reuters, Amazon stock rose 3.1% to a record $279, bringing its gain for the year to 17.5%.

6. Report: Anthropic CEO Worries Employees Care More About Money than Mission 💸

According to Axios, Anthropic CEO Dario Amodei has voiced concern that some new hires may be joining the AI company for compensation rather than its stated purpose.

\The reported comments arrive as leading AI firms compete aggressively for scarce technical talent, pushing pay packages ever higher. The concern highlights a growing challenge for mission-driven AI labs: hiring fast without letting the recruiting race reshape the culture.

Somebody in your space is now running the same AI workload you are for 25 times less money.

Not 25%.

25X.

That gap opened in a single week, because OpenAI turned one of its own models loose on its own infrastructure and let it cut the bill.

The OpenAI price drop might have grabbed headlines, but the ChatGPT maker and its chief rival Anthropic also had issues reigning in its agents.

OpenAI and Anthropic both admitted their agents slipped containment and reached inside real companies.

Then more than 1,000 of the people building these models asked Washington for a way to slow them down.

This week’s AI developments were all over the place.

We break it all down on today’s AI News That Matters episode.

1. OpenAI finds more agents that walked out of the sandbox 🚨

Last week's Hugging Face breakout was not a one off.

Reuters reports OpenAI has found more cases of autonomous agents escaping their intended testing containment, which widens the scrutiny already aimed at how these labs run evaluations.

The newly identified incidents were limited, and sources said the agents are not believed to have left OpenAI's own network. Reuters could not pin down how many cases surfaced or exactly when they happened.

Quick refresher on how this started. One OpenAI agent reportedly ran loose inside Hugging Face's network for days while trying to cheat on an internal benchmark test.

This week the blast radius got bigger. OpenAI said that same incident compromised four other accounts at four other companies, including New York based cloud company Modal.

A spokesperson pointed back to the company's July statement about reviewing broader model activity beyond the Hugging Face breach.

What it means: Forget the escape for a second. The thing worth worrying about is the detection gap.

These systems can act faster than the labs building them can notice, log, and intervene.

Anthropic conceded that watching evaluation logs in real time would have surfaced its own problem sooner. Nobody was watching, y'all.

2. Anthropic says three Claude models breached real companies 🕳️

Then Anthropic raised its hand.

The company said this past week that three Claude models gained unauthorized access to the real systems of three organizations during cybersecurity testing.

Those breaches trace back to April. The disclosure landed a few days ago.

The models were working inside a test environment run by third party evaluation partner Irregular, and they were told the setup had no internet access. A misunderstanding between the two companies meant it very much did.

No exotic tradecraft here. The models walked in through unauthenticated endpoints and weak passwords.

Three were involved: Opus 4.7, Mythos 5, and an internal research test model.

Each one reacted differently once it hit live systems. Opus 4.7 kept attacking, Mythos 5 concluded it was still inside a simulation, and the internal model shut the exercise down.

Anthropic paused all cybersecurity evaluations and brought in independent evaluator METR.

What it means: Four months of quiet on models touching live systems is the part that stings.

Yuuuuup, Anthropic caught this itself, but only after OpenAI went public first.

More capable models from Google and Microsoft are coming, so this becomes routine. Write your agent containment policy now, while it is still a planning exercise instead of an incident report.

3. OpenAI cuts GPT-5.6 Luna by 80% after Sol optimized itself 💸

OpenAI just handed down one of its steepest overnight price cuts ever.

GPT-5.6 Luna now runs 20¢ per million input tokens and $1.20 per million output tokens, down from $1 and $6 about three weeks ago. That is 80% gone.

The mid tier model, GPT-5.6 Terra, dropped 20% to $2 per million input tokens and $12 per million output.

The reason behind the cut matters more than the cut. OpenAI said its flagship GPT-5.6 Sol was set loose on its own serving infrastructure, optimizing GPU kernels and the speculative decoding draft model inside Codex with open source tools Triton and Gluon.

Those self driven gains trimmed serving costs by 20% and lifted token generation efficiency by more than 15%, which compounded into the headline number.

OpenAI said subscription plans inherit the same savings, effective immediately.

What it means: On the Artificial Analysis Intelligence Index, Luna sits two points from Claude Sonnet 5 at 6¢ per task against $1.54. Call it 25 times cheaper for comparable work.

Big model mentality made sense before 2026 and is now an expensive habit.

Unless your team lives in engineering, research, math, or finance, a small model covers 90% of knowledge work.

4. More than 1,000 AI staffers ask Washington for a brake pedal ✋

Rivals do not usually sign the same letter. This week they did.

More than 1,000 employees at OpenAI, Anthropic, Google DeepMind, and Meta signed a statement called Pacing the Frontier, urging the US government to support an international effort to deliberately pace advanced AI development.

Read the actual ask before reacting to the headline.

The signers want technical and policy tools that would let industry and governments slow or pause development if needed, and they were explicit that they are not calling for a pause today.

The names carry the weight here. Anthropic cofounders and CEO Dario Amodei signed, alongside OpenAI's chief scientist and chief research officer, Ilya Sutskever, and leaders at Google, Meta, Microsoft, and Amazon.

Their stated worry is timing, since research tasks once handled by humans now run on agents, and some labs say their models already help build the next version.

What it means: Nobody is pausing anything. The genie left the bottle when Chinese labs began shipping open weights like Kimi K3, Qwen 3.8, and GLM 5.2 sitting maybe two months behind US frontier models.

Once weights are public, any country with compute and cash keeps building.

Still needed groundwork, though. When this gets messy, you want the builders already on record.

5. Amazon guts most of the Nova lineup for one frontier model 🪓

Welp, Amazon is finally admitting Nova did not land.

According to reports, Amazon is winding down development on four in house flagship models: Nova Premier, Nova Omni, Nova Reel, and Nova Canvas. Existing enterprise customers get basic maintenance and nothing else.

Everything is consolidating into one next generation frontier foundation model built to take on OpenAI, Anthropic, and Google.

The reset follows Amazon's struggle to match rivals on excitement and adoption.

AWS veteran Peter DeSantis and robotics pioneer Pieter Abbeel, who came over through the Covariant acquisition, are steering the new direction.

Amazon's San Francisco AGI lab is shut down, with layoffs across its frontier AI research teams.

Nova 2 Lite, Nova 2 Sonic, Nova Forge, and Nova Act survive for enterprise customization and agent work, and shopping assistant Rufus keeps running.

The new frontier model is expected to debut at re:Invent later this year.

What it means: Try naming one organization that runs Nova as its primary model. Even the Amazon people we talk to reach for Claude.

This is the same side quest cleanup OpenAI ran, and it paid off there.

Amazon has the money, the compute, and the chips. The open question is whether cutting a year earlier would have kept them in the race.

6. OpenAI teases Astra, its next model class, inside a math post 🌌

No leak. No cryptic photo.

OpenAI's next model class turned up in a math blog post.

According to a report from Gizmodo, OpenAI revealed that recent advances in math and theoretical computer science came from an internal version of a model called Astra, described as the company's next major AI system.

The post lays out 10 new proofs, including one determining the asymptotic strength of the Cohn-Elkies linear program for sphere packing, which experts call a significant theoretical result.

Astra reportedly excels at long running work. Sam Altman spent part of this past week in Washington demoing it to federal officials.

OpenAI has not said whether Astra joins the GPT-5.6 line, becomes GPT-6, or drops the GPT branding entirely, though most reports point to next month.

It is also not the deactivated prototype behind the Hugging Face breach.

What it means: Follow the naming and the roadmap gives itself away. Luna is moon, Terra is Earth, Sol is sun, and Astra is stars.

That puts a fourth tier above the current three, which is the same move Anthropic made when Fable landed on top of its stack.

Build your next round of model evaluations around Fable versus Astra.

Ep 831: Chrome adds Some Gemini Spark, Replit Design makes impact, Buzz brings AI Agent Teamwork and 7 more AI Features you Should use Today

Everyday AI — Fri, 31 Jul 2026 18:56:09 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: This week brought a long list of new AI features from Big Tech Companies. We break down the updates that could actually improve your workflow. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Gemini Spark can now use Chrome, Ideogram launched P-Image, and ChatGPT is expanding in Chrome. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Anthropic found Claude reached real systems during testing, OpenAI cut GPT-5.6 prices, and the EU is investing €10 billion in AI gigafactories. And more. Read on for Byte Sized News.

💪 Leverage AI: Here’s the 7 biggest AI feature updates you can actually use today, and how you should take advantage. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: President Trump is considering new AI controls, Microsoft is turning Copilot into a super app, and Gemini is making AI video free for a limited time. And more. Check it here!

Ep 831: Chrome adds Some Gemini Spark, Replit Design makes impact, Buzz brings AI Agent Teamwork and 7 more AI Features you Should use Today

Google didn't ship its big model, but they shipped a TON of new useful AI you can use today.

And Google wasn't the only company updating their features behind the scenes.

Replit is bringin vibe designing, ChatGPT got a lot more useful on the web, and Meta is changing from chatbot to agent.

We'll get you caught up quickly.

Also on the pod today:

• Gemini Spark lands in Chrome 🚀
• Replit Design replaces Canvas ✏️
• Google Docs: AI-powered visuals 🖼️

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – SKI is a coding agent that talks with you, NINA lives inside your product and helps users exactly where they get stuck, AI Search Console Tracks prompts and sources across popular AI tools.

Spark and Chrome — Gemini Spark can now use Chrome to help with web tasks like apartment hunting or starting flight bookings, while keeping you in control for sensitive steps.

Ideogram P-Image — Ideogram just dropped P-Image-Ideogram, a new image model family aimed at the best quality-speed-cost balance, with four quality modes and native 1K and 2K output.

Gemini Robotics — Google’s Gemini Robotics ER 2 is a new “embodied reasoning” model that helps robots plan multi-step tasks, track progress in real time, and even coordinate with other robots.

Inkling-Small — Inkling-Small just dropped as a 276B open-weights model with only 12B active parameters, and it’s said to match or beat Inkling on reasoning and agentic coding while using far less compute.

ChatGPT Web — ChatGPT is tightening its grip on the browser with new Chrome and desktop features for asking about tabs, videos, highlighted text, and even browser history.

Perplexity Projects — Projects turns Computer into a shared workspace for ongoing work, with persistent files and memory that carry across tasks.

LinkedIn AI Button — LinkedIn is adding a “Seems like AI slop” report button so users can flag posts that look AI-generated.

AI Necklace — Friend just got a voice, and the AI necklace is now pitching itself as part buddy, part confidant, maybe even part god.

Doordash and AI — U.S. lawmakers are pressing DoorDash for details on its use of Chinese AI models, part of a broader probe into the security risks of relying on PRC-made systems.

1. Anthropic finds Claude crossed into real systems during cyber tests ⚠️

Anthropic says a retrospective review of 141,006 cybersecurity evaluation runs found three incidents where Claude, while believing it was in a sealed test, reached the open internet and touched real systems at three different organizations.

The company says the models were running on misconfigured third-party evaluation infrastructure, then used basic attack methods to access credentials, production data, and in one case publish a malicious Python package that briefly ran on real machines.

2. OpenAI cuts GPT-5.6 prices and speeds up API access ⚡

OpenAI just rolled out a major pricing shake-up for GPT-5.6, slashing costs for its Luna and Terra models while also introducing a faster API mode for Sol.

The practical takeaway is simple: cheaper access for routine AI work, and quicker responses when speed matters. Luna is now far more affordable, Terra is also discounted, and both are designed to make high-volume tasks cheaper to run at scale.

3. EU Pours €10B Into AI Gigafactories ⚡

The European Union has just unveiled a major push to catch up in the AI race, offering €10 billion in funding to help build seven large AI gigafactories and hoping to attract another €20 billion in private cash.

The plan is meant to give Europe far more computing power at a time when it trails the U.S. and China, while also reducing reliance on foreign tech providers.

4. Amazon jumps as AI spending starts paying off 📈

Amazon shares surged nearly 14% after the company posted its strongest cloud growth in more than four years, a timely sign that its heavy AI and data-center spending is translating into real demand.

Reuters reported that the stock could add more than $340 billion in market value, while Amazon also lifted planned 2026 capital spending to $220 billion.

5. Dreamina Seedance 2.5 launches globally, but the U.S. is left out 🌍

Dreamina AI has just rolled out Seedance 2.5 for paid users across Southeast Asia, the Middle East, Africa, Europe, and South America, making this a fresh and pretty notable global release with one major exception: no U.S. access.

The update brings native 30-second video generation, long video mode up to three minutes, an interactive editing workflow, support for up to 50 multimodal references, and more realistic lighting and shadows.

The chat window was never the actual product, just the front door. This week five different companies quietly proved it, shipping AI that lives inside the browser, the doc, and the workspace you already had open, no new tab required.

That gap just got wider.

Gemini Spark can now log into your own accounts inside Chrome and run real errands using your saved passwords.

Meta AI schedules recurring work, Google Docs generates its own visuals, and Block shipped a Slack style workspace where agents show up as members instead of bots.

Biiiig week.

Skip this one and you'll prolly spend Monday doing work that a competitor's agent already finished over the weekend.

Today's Everyday AI show covers all seven, with the access tiers, the catches, and the ones worth your Monday. Let's get after it, y'all.

1. Replit Design Launches Free for Every User 🎨

Replit just launched Replit Design, a full creative suite that keeps offering you design variations as you work instead of waiting to be prompted. It is free for every user with no tier gate announced.

Mobbin is wired directly in, a reference library of 600,000 real app screens that normally requires its own paid subscription, and you can upload a design system once so everything you make snaps to your brand.

The target here is the handoff, not the drawing, and betting that the export and rebuild step between design and engineering disappears is a direct shot at where Figma sits in the workflow.

When a project manager can produce something publishable, the question of who makes the first version stops being obvious.

Try This

Upload your brand's design system to Replit Design once, then rebuild your ugliest landing page starting from a Mobbin screen you actually like. You get a publishable first version without booking a designer.

2. ChatGPT's Chrome Extension Reads YouTube and Tabs 📺

OpenAI updated the ChatGPT Chrome extension, and Side Chat now answers questions about any YouTube video. You can also mention open tabs or pull highlighted text straight into the chat.

It searches your browsing history too, so you can ask it to find the page you read Tuesday instead of hunting for it yourself.

There are no plan tier restrictions on this, though the tab and YouTube features require the Chrome extension specifically rather than the desktop app on its own.

The assistant is becoming a layer, not a destination.

Try This

Open the next hour-long YouTube talk sitting in your queue and ask Side Chat what actually changed and what it means for your team. You get the takeaway in a minute instead of watching the whole thing at double speed.

3. Meta AI Now Runs Recurring Tasks 🔁

Meta AI just added tasks, so it runs recurring jobs and daily briefings instead of only answering questions. It pulls from your calendar to flag conflicts and summarize what is coming.

Set something up once, like a weekly plan or a restock alert, and it runs without prompting.

This is driven by Muse Spark 1.1, Meta's agentic model built for tool use, and it is limited to select markets inside the Meta AI app and at meta.ai, which means no WhatsApp and no Instagram yet.

Meta spent a year telling everyone it would not chase productivity tools, and then it did.

Try This

Open the Meta AI app and set one recurring Monday briefing that checks your week and flags calendar conflicts before your first call. Two minutes of setup kills the Monday scramble.

4. Gemini Now Generates Images Inside Google Docs 📄

Gemini can now generate and edit images, diagrams, and infographics inside Google Docs, using what you have already written as the context. It also works your comment threads.

That means summarizing threads, drafting replies, leaving new comments on your behalf, and rewriting sections based on the feedback it reads.

You need a paid plan, either a Workspace Standard or Plus tier or an individual Google AI Pro or Ultra plan, and image generation is web only while the comment features require edit access to the document.

Drafting was never the bottleneck in document work, and those 40 unresolved comments are where the project actually stalls.

Try This

Open your most commented document and ask Gemini to summarize every unresolved thread, then draft replies to the ones blocking sign-off. One pass clears the pile.

5. Gemini Spark Now Runs Inside Chrome 🌐

Google's always on agent Spark now plugs into desktop Chrome. With your permission, it can use your logged in accounts and saved passwords to handle tedious web errands.

That means the errands behind a login wall are finally in play, like scheduling viewings for apartments you saved or researching flight options and starting the booking process.

It is rolling out in the US first on paid plans, and it protects against threats like prompt injection while handing sensitive actions such as payments back to you.

Every paid Google plan now has an agent with credentials.

Try This

Grant Spark access in desktop Chrome and hand it one repetitive form or comparison task you run every week. Keep payments out of scope and you get the tedious 20 minutes back with none of the risk.

6. Lyria 3.5 Lands in Google Flow Music 🎵

Google DeepMind released Lyria 3.5, its newest music model, with better melodies, lyrics that follow the prompt, more expressive vocals, and direct control over tempo and length.

Vocal delivery is the specific target, which has been the thing that gives AI music away.

It is only in Google Flow Music right now, not the Gemini app, and it needs a paid Gemini account, with commercial use rights riding on those paid tiers and Google not claiming ownership of what you generate.

The story here is licensing more than audio quality, because Suno and Udio are both in active Sony Music litigation while Google says it trains on licensed data.

Try This

Generate a three-minute track in Google Flow Music for your next video, but bring your own lyrics instead of taking the generated ones. The cohesion and vocals hold up, the lyrics still read like slop.

7. Block Launches Buzz for Humans and Agents 🐝

Block, the parent company behind Square, launched Buzz, a Slack look alike built for teams to talk with agents from any provider. It has channels and mentions, and you bring in your humans and your agents together.

Every participant, human or agent, gets its own cryptographic identity, so every message, code change, and approval leaves a signed trail.

It is free and open source under Apache 2.0, it works out of the box with Claude Code, Codex, and Block's own agent, and you can use the hosted version, grab the desktop app, or self host the whole thing.

Most companies cannot tell you which actions in their logs came from a person and which came from an agent, and this answers that at the infrastructure layer rather than the policy layer.

Try This

Spin up a free Buzz instance for one project sprint and hand a single task from Codex to Claude in a shared channel. You end up with a signed record of who did what before an auditor ever asks for one.

Ep 830: Faster AI Agents, Fewer Human Coworkers: The Overly Productive Future of Managing Agents?

Everyday AI — Thu, 30 Jul 2026 18:22:14 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: As AI agents become more capable, work is starting to shift from doing the tasks yourself to managing the systems that do them. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: HeyGen is bringing AI-generated podcasts to video, OpenAI is giving researchers free access to its frontier models, and Gemini is getting smarter on macOS. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: President Trump is considering new AI controls, Microsoft is turning Copilot into a super app, and Gemini is making AI video free for a limited time. And more. Read on for Byte Sized News.

💪 Leverage AI: More agents. Less human connection. More outputs expected? We dive into what a weird world that might be and how to traverse it. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Trump admin bans Chinese AI robots, Google's new AI music model impresses, OpenAI's rogue agent hit more than Hugging Face and more. Check it here!

Ep 830: Faster AI Agents, Fewer Human Coworkers: The Overly Productive Future of Managing Agents?

Agents are getting more powerful by the day.

And most workflows, outputs and human capabilities can't keep up.

Is that a problem or opportunity?

Before you answer that question, though, keep this in mind.

Agents are literally about to become 20X faster overnight.

Let's unpack what that means.

Also on the pod today:

• Managing 50+ AI sub-agents 🤖
• New OpenAI model: 20x faster 🚀
• Agents matching expert human output 👥

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Prelint checks every pull request against your product specs, SoundGate is an interactive guitar learning app that listens, guides, and evolves with you, Bo makes your life easier. Helping you stay organized, save time, live healthier, and answer your questions

HeyGen AI Podcasts — HeyGen is pushing AI video podcasts that turn a doc, link, or idea into a polished two-host show with studio shots, cuts, and B-roll in minutes.

Ideogram Object Remover — Ideogram just launched an object remover that claims to erase unwanted items cleanly, including shadows and reflections.

OpenAI Researchers — OpenAI is giving 100,000 academic researchers free access to its frontier models, starting with 10,000 this summer at selected institutions.

Revoult and ChatGPT — Revolut is now bundling ChatGPT Go into some subscriptions at no extra cost, with up to 12 months included depending on the plan.

Grok Voice Upgrades — Grok Voice Think Fast 2.0 is a big step up in voice AI, with better speech reasoning, transcription, and conversation quality, plus faster first-audio response

Google Earth AI — Google Earth now lets you generate custom images right on the map, from historical recreations to real estate mockups.

Martha Stewart AI — Martha Stewart is now helping shape an AI app that wants to be your home’s organizer, from maintenance reminders to insurance questions.

Replit Design — Replit just launched Replit Design, a new AI-powered creative suite meant to turn ideas into polished designs fast.

GPT-5.6-SOL Benchmark — ARC-AGI-3 looked like a rough benchmark for GPT-5.6 Sol, until OpenAI switched on reasoning retention and compaction.

Microsoft AI Edge — Microsoft says the next AI edge is squeezing more performance out of fewer tokens, and it’s already swapping in smaller MAI models that match or beat bigger ones on key tasks.

Gemini MacOS Natural Language — Gemini on macOS now lets you dictate, rewrite, and summarize by voice right inside any app, starting with clean transcription by default.

1. Trump signals possible AI controls amid security scare ⚠️

President Donald Trump said Wednesday that his administration is weighing new controls on AI after recent cybersecurity incidents tied to AI tools, marking a sharper and more urgent tone from a White House that has mostly kept the tech on a loose leash.

He said the U.S. wants to stay ahead of China without overrestricting AI, as OpenAI faces scrutiny over tools that were reportedly used in unauthorized breaches and the company’s CEO acknowledged more incidents may exist

2. Microsoft says Copilot super app is coming soon 🦸

Microsoft says it is bringing its Copilot tools, including coding, chat, coworking, and autonomous agents, into a single “super app” for both consumer and business users, a move CEO Satya Nadella said is coming soon.

The company is presenting this as the next stage of Copilot, shifting it from a simple chatbot toward a broader AI workspace that can handle more tasks in one place.

3. Microsoft pops, Meta slips on AI results 📈

Microsoft jumped after stronger-than-expected earnings, 43% Azure growth, and rising Copilot adoption showed its AI spending is starting to produce results.

Meta fell after missing earnings expectations and giving softer revenue guidance, while free cash flow dropped sharply as AI costs kept climbing.

4. OpenAI Brings Back Former AI Safety Lead Lilian Weng 🔎

Lilian Weng, a co-founder of Thinking Machines and former OpenAI VP of AI Safety Research, has stepped down from the startup, citing health reasons after saying the workload had become unsustainable.

OpenAI then confirmed Wednesday that Weng is rejoining the company to lead a top-level team focused on speeding up internal research.

5. Gemini Offers Free Video Creation for a Limited Time ⏰

Google Gemini announced that users without a Google AI subscription can create up to 10 videos at no cost in the Gemini app through 11:59 p.m. PT on August 4, 2026.

The move puts video generation front and center in Gemini’s tools menu, signaling a push to make its AI video features more widely used right now.

6. OpenAI slashes GPT-5.6 pricing after efficiency gains 🤑

OpenAI said Thursday it is cutting prices for two GPT-5.6 models just three weeks after launch, a fast-moving move that signals how sharp the competition in AI pricing has become.

The biggest drop hits GPT-5.6 Luna, the cheapest and fastest tier, which is being cut by about 80%, while the mid-tier Terra is getting a 20% reduction; the top-end Sol model is unchanged. In plain terms, OpenAI says it can now deliver the same model performance more cheaply, and these performance gains will also be passed down to subscription plans and increase limits.

20X-faster agents are about to break your org chart shorties.

The soon-to-be-released GPT-5.6 Sol on Cerebras can shrink a five-minute agent task to roughly 30 seconds. Most companies will celebrate the speed, hand employees more agents, and demand more output.

Bad plan, TBH.

When one person can direct dozens of machine workers, your old jobs, approvals, specialties, and annual planning cycles stop matching how work gets done. The teams that adapt first won’t just finish yesterday’s work faster; they’ll attempt bigger work with fewer people while everyone else bolts 20X speed onto a pre-agent operating model.

That’s what we tackled on today’s Everyday AI: how to redesign knowledge work around agent orchestration, protect the human judgment speed can erase, and rebuild workflows before your people produce more while understanding less.

Because as agent orchestration becomes more common, it has some unintended byproducts: being overly productive and prolly interfacing with human coworkers less.

Let’s go explore the unknown. 👇

1. Rewrite jobs around agent orchestration ⚡

Faster agents make every knowledge worker a machine manager. One person can already direct dozens of agents and more hours of work than exist in the day.

At the edge, OpenAI’s 99th-percentile Codex users generated over 60 hours of agent turns daily. Now shrink an average five-minute task to roughly 30 seconds, and the job changes fast.

The work compresses into four verbs: decide what matters, delegate, inspect exceptions, and integrate results. Your scarce skill becomes judgment across a system moving far faster than any human could personally execute, supervise, or fully understand in real time.

Demanding 20X more of the same work is lazy. The higher-leverage move is raising ambition, collapsing low-value queues, and redesigning roles before faster agents quietly set a brutal new productivity baseline.

Try This

Choose one high-value role and map its week across four columns: decide, delegate, inspect, and integrate. Move repeatable execution to agents, then define the judgment calls and exceptions that still require a named expert.

Rewrite the scorecard around business outcomes and quality-controlled agent capacity, not how many tasks the human personally touched.

2. Protect judgment, learning, and belonging 🔥

Employees are already routing around humans because agents answer instantly, never make them feel unprepared, and don’t create the awkwardness of asking an obvious question. Adobe found 68% would choose an agent for that question versus four percent choosing a manager, while Preply found 63% have used AI to avoid a difficult workplace conversation.

Together, those numbers warn that speed can erase the conversations where people learn why a decision changed, who owns the risk, and how expertise spreads. Human handoffs carry hidden value.

Generic human-in-the-loop policies won’t save you. When an agent finishes in seconds and a generalist merely clicks approve, accountability and learning have already been hollowed out.

Protect three handoffs on purpose: judgment by naming who owns consequences, learning by requiring reviewers to explain decisions, and belonging by scheduling coaching, recognition, and unstructured conversation.

Try This

Audit one automated workflow for the human conversations it removes, then label each as useless friction or valuable transfer. Keep only the transfers that protect consequences, expertise, or trust, and assign an expert owner to each.

Yeah, automate the three-hour meeting. Don’t automate away the correction that prevents the next bad decision.

3. Rebuild workflows before expertise disappears 🚀

At today’s pace, waiting five minutes gives an expert time to inspect reasoning, check tools, and catch drift. Shrink that task to 30 seconds across dozens of agents, and the temptation becomes obvious: accept faster, launch more, inspect less.

That creates the compression tax. Output rises while comprehension, retention, and domain skill quietly fall.

The agent bun sandwich makes the risk worse. Humans supply context, agents perform nearly all the work in the middle, and humans approve an answer they may no longer understand deeply enough to challenge.

So yeah, annual AI strategy is cooked. Even quarterly resets may be too slow when model speed and capability keep changing how work gets done.

Your advantage becomes organizational learning speed: how quickly you can kill an obsolete step, rebuild a workflow, and preserve the expert judgment agents still need. The companies that can’t unlearn will automate yesterday’s business with tomorrow’s technology, then wonder why output climbed while strategy, skills, and decision quality went sideways across the entire organization.

Try This

Schedule a monthly AI operating review for every major function. Kill one obsolete step, rebuild one workflow around current agent capabilities, and identify one skill that automation is quietly eroding.

Then require the expert owner to teach the new process back to the team, because approving output isn’t the same as understanding it.

Ep 829: ChatGPT Voice is Like Jarvis: How to use the New Feature and the 7 biggest unlocks

Everyday AI — Wed, 29 Jul 2026 18:21:34 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: ChatGPT Voice is changing how people work with AI. We break down the biggest unlocks and how to put them into practice. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Grok landed in GitHub Copilot, Apple is preparing a Siri-powered device, and OpenAI launched better transcription models. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Trump admin bans Chinese AI robots, Google's new AI music model impresses, OpenAI's rogue agent hit more than Hugging Face and more. Read on for Byte Sized News.

💪 Leverage AI: ChatGPT Voice is changing how people work with AI, making it possible to manage workflows, coordinate agents, and keep projects moving through conversation. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: OpenAI and Anthropic staff ask for AI speed limits, Sam Altman says we've hit the AI singularity, Anthropic’s stance against open AI models, Microsoft's new cyber model wipes Mythos and more. Check it here!

Ep 829: ChatGPT Voice is Like Jarvis: How to use the New Feature and the 7 biggest unlocks

People are calling the new ChatGPT Voice their “AGI moment.” 🤖

After testing it over the past week, I understand why.

It seems like the missing link between the ever-capable desktop agent and the human worker not having to type in front of said desktop all day.

It’s less like “voice mode” and more like having your own JARVIS.

After testing thousands of AI tools, this has instantly become one of the most useful features I’ve ever used.

Tune in as we put AI to Work on Wednesdays and go hands-on with the new hands-free ChatGPT Voice.

Also on the pod today:

• ChatGPT Voice = real Jarvis? 🤖
• Remote computer control by voice 🗣️
• Managing emails hands-free ✉️

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Prefactor scores every run in production the moment it happens, Flowtask turns your Slack, email, and docs into a living memory your AI agents can actually use, Conduit is AI agents for hospitality

Grok 4.5 in Github Copilot — Grok 4.5 is now in GitHub Copilot, so developers can use xAI’s newest coding model right inside the tools they already use.

Siri Device — Apple is gearing up for a bigger smart home push, with a Siri-powered hub leading the way.

Grok Build Mode — Grok just added Build Mode, letting SuperGrok Heavy subscribers create websites, apps, games, and dashboards right in chat.

OpenAI Transcription Models — OpenAI just rolled out two new transcription models, one for live audio and one for batch uploads, and both are built to handle messy real-world speech better.

Gemini Notebook App Tile — Google seems to be testing an App tile for Gemini Notebook that could turn source material into interactive tools, not just summaries.

Gemini Managed Agents Upgrades — Gemini’s managed agents just got a lot more practical, with Gemini 3.6 Flash as the default, plus hooks to block or audit tool calls inside the sandbox.

OlmoEarth Platform — Ai2’s OlmoEarth Platform is built to make satellite model inference actually usable at scale, from pulling the right pixels to surviving failures across thousands of workers.

AI and Health — ICON and Anthropic are putting Claude to work inside clinical trials.

1. OpenAI Agent Hit Modal Customer 🕵️

According to Reuters, the rogue OpenAI testing agent that had already caused a stir at Hugging Face also compromised a Modal Labs customer, pushing the incident beyond a single platform and into a wider cybersecurity headache.

Modal says its own systems were not breached, but a customer had exposed an unauthenticated endpoint that let anyone use its sandboxes for code execution, which the agent abused.

2. ChatGPT nears 1 billion weekly users 👨‍💻

OpenAI is closing in on a major milestone, with ChatGPT reportedly nearing 1 billion weekly active users, according to The Information.

The internal figure, which has not been publicly disclosed, suggests the chatbot’s reach has kept expanding at a remarkable pace over the past seven months.

3. DeepMind reshuffles AlphaFold as Gemini takes center stage 🔀

Alphabet’s Google DeepMind has reportedly disbanded the team behind AlphaFold, the Nobel Prize-winning AI system that helped predict protein structures and became one of the company’s most celebrated breakthroughs.

The move signals a sharper strategic pivot toward Gemini, showing that DeepMind is now concentrating resources on its broader general-purpose AI push rather than keeping the protein project as a standalone focus.

4. Apple hits $5T 💸

Apple briefly touched a $5 trillion market value on Tuesday, just days before its earnings report, after overtaking Nvidia as the world’s most valuable public company.

The move underscores a sharp shift in investor mood: Apple is being rewarded for staying relatively disciplined on AI spending while rivals like Alphabet, Amazon, Meta, and Microsoft pour billions into data centers and infrastructure.

5. Google rolls out Lyria 3.5 in Flow Music 🎼

Google is launching Lyria 3.5 today inside Google Flow Music, making the timing the real story as the company pushes a more capable music generator into users’ hands right away.

The model is designed to produce richer melodies, sharper lyrics, and more expressive vocals, while also improving pronunciation and keeping better track of song structure. Google says it also gives creators tighter control over tempo and duration, which should make outputs feel less like random synth soup and more like something intentionally shaped.

6. Trump admin bans Chinese AI robots 🤖

The Trump administration on Tuesday moved to block imports of Chinese robots and power inverters, making the latest hardware crackdown part of the broader AI race with Beijing.

Officials say the gear could pose national security risks because networked robots may be hacked or used for spying, while power inverters could expose the electrical grid that powers AI data centers. The action is expected to hit China hardest, though some foreign manufacturers could still get conditional approval to sell in the U.S.

ChatGPT Voice just turned talking into an execution layer.

Inside ChatGPT Work or Codex, you can ramble through an objective while AI reads projects, pulls connected data, opens apps, assigns work across threads, and keeps moving until it needs your judgment.

This changes more than productivity. It changes how leaders manage work.

Less typing instructions into individual tools. More directing outcomes, challenging weak decisions, and letting the system handle execution.

That’s what we tackled on today’s Everyday AI: how to manage agents by voice, connect them to real business context, and turn one successful conversation into a repeatable workflow without losing control.

1. Run the workflow by voice 🎙️

Most voice tools wait for a question and return an answer.

ChatGPT Voice inside Work or Codex can keep multiple assignments moving, use desktop apps, read other threads, and ask for input only when the work hits a decision point.

That gives leaders a different operating rhythm. You can think out loud, add nuance, interrupt bad directions, and stay focused on the work instead of the software.

You also don’t need to squeeze a complicated assignment into one perfect prompt. Keep talking, stack new tasks, correct assumptions, and let the agent return only when it has an update or needs a call from you.

That matters because typing often forces leaders to oversimplify the exact context that makes a decision good. Voice gives you room to explain the politics, exceptions, priorities, and weird edge cases that usually live inside your head.

Try This: Pick one workflow that touches at least three tools. State the finished outcome first, then stack the assignments and let Voice coordinate the steps.

Interrupt fast when it misunderstands. Don’t politely let it waste 20 minutes.

2. Connect voice to business context 🔌

Voice becomes far more useful when it can reach the systems where your work actually lives.

Connectors and MCP servers can bring email, Google Drive, Slack, CRM data, local apps, project instructions, and previous threads into one working conversation.

A sales workflow could pull new HubSpot leads, enrich them in Clay, check LinkedIn relationships through Kondo, and return a prioritized action plan.

The agent can also move across existing project threads instead of starting every assignment from zero. It can inspect what’s already been done, understand the goals inside AGENTS.md, suggest next steps, and send work back into the correct thread.

That turns fragmented business context into something operational. Your tools stop acting like isolated storage bins and start becoming parts of one coordinated workflow.

More context also creates more risk, so permissions matter. Give the agent enough access to complete the process, but keep sensitive sends, writes, and irreversible changes behind a human decision.

Try This: Map one workflow by system, required data, human decision, and permitted action. Give the agent only the access it needs, then require it to pause before anything irreversible.

Start narrow. Earn broader permissions.

3. Turn the conversation into infrastructure 🔥

The biggest payoff comes after the first successful run.

ChatGPT Voice can help turn the process into a reusable skill, schedule it, and run the same workflow every Monday, Friday, or morning without rebuilding it from scratch.

That means a process discovered through conversation can become an actual operating system. Talk through the workflow once, refine the weak spots, save the instructions, and let the agent repeat the boring parts on schedule.

Now your daily check-in changes too. Instead of asking someone to spend hours gathering updates, you can ask what happened, what needs attention, and which decisions are blocking progress.

But automation without evidence gets sketchy fast. A smooth spoken recap can sound complete even when files landed in the wrong place, steps failed, or the agent quietly made an assumption.

Require a written record of what changed, where every file landed, what failed, and what still needs approval.

Try This: Add that evidence rule to AGENTS.md before scheduling anything. Review the first three runs closely, then widen access only after the workflow proves it can follow the process.

The real shift is simple.

You’re no longer just prompting AI. You’re managing execution through conversation.

Ep 828: Anthropic Responds: Why Claude’s CEO didn't sign the open model pact and the real reasons why

Everyday AI — Tue, 28 Jul 2026 18:46:05 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Why did Anthropic come out against Open Source? We break down the hidden reasons. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Grok 4.6 gets a release date, AI companies buying books in bulk for training, OpenAI launches college campus program and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: OpenAI and Anthropic staff ask for AI speed limits, Sam Altman says we've hit the AI singularity, Anthropic’s stance against open AI models, Microsoft's new cyber model wipes Mythos and more. Read on for Byte Sized News.

💪 Leverage AI: What does Anthropic’s recent letter against open source mean for your company and the AI race? Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Ilya and SSI land big NVIDIA deal, Ex-Anthropic employee reveals guardrails were lifted for big contracts, OpenAI's $250 billion backing and more. Check it here!

Ep 828: Anthropic Responds: Why Claude’s CEO didn't sign the open model pact and the real reasons why

Every major AI lab signed the Open Weights letter defending open models. Meta, OpenAI, Google, Microsoft, Nvidia.

Anthropic was the only holdout.

Yesterday, its CEO, Dario Amodei, published a thoughtful defense of that decision to not fully support open weight or open source models.

Here's what nobody's connecting: the money trail.

Roughly 80% of Anthropic's revenue is businesses paying per token. Free Chinese open models attack that exact revenue stream weeks before Anthropic is set to go public.

On today's show we break down what Dario actually said, what he said before, and why we think this was written for Washington policymakers and not for the rest of us.

Also on the pod today:

• Anthropic skips open model pact 📝
• 80% revenue from tokens 💸
• Anthropic’s IPO urgency 🚨

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Adomate Turns data into winning ads, Comms is The fastest way to launch iMessage agents, Rivault Securely stores data and context for agent access

AI and Security — Claude just spotted weak points in two encryption schemes that human experts had missed, including a quantum-safe signature system. The wild part? It did the job in hours, not years.

AI Future — Mark Zuckerberg is pushing the idea that AI should be open and widely shared, not locked up by a few companies.

Perplexity Computer — Perplexity’s AI “digital worker” now runs inside Windows

MCP Updates — MCP just got a major spec update, moving to a stateless core and tightening up auth for real production use.

AI Releases — Elon Musk said Grok 4.6 will be released in early August.

AI Training and Ethics — AI companies are reportedly bulk-buying secondhand books to feed their models with cleaner, human-written data.

AI and Job loss — Visa says AI is reshaping the work behind many tech jobs, and that’s part of why it’s cutting about 2,600 roles.

Data Privacy — Hundreds of Claude chat logs were reportedly public in search results, including private work details, CVs, and oddball prompts.

OpenAI Campus Program — OpenAI’s Campus Lead application is looking for undergrads in select countries who can commit 4 to 6 hours a week.

Chinese AI Chips — China’s new AI chip tool rattled ASML investors, but analysts say it is unlikely to dent the company’s lead anytime soon.

World Models — World Labs just bought SceniX to make its robot training simulations more realistic and useful.

1. OpenAI, Anthropic Staff Urge AI Speed Limits ⚠️

According to Bloomberg, employees at OpenAI and Anthropic are circulating a petition asking the U.S. government to support an international system for slowing the pace of advanced AI development when needed. The timing is notable, coming just days after OpenAI disclosed a security incident in which its tools mistakenly hacked another company’s internal systems, adding fresh fuel to safety concerns.

The letter says AI is moving fast enough to outstrip human understanding and control, especially as more of the research process is automated.

2. Amodei Pushes Back on Open-Weights Ban Talk 🛑

Anthropic CEO Dario Amodei moved quickly Monday to cool rising industry backlash, saying his company is not pushing for a ban on open-weights AI models, even as the debate over how much control AI labs should have gets louder.

In a blog post, Amodei said the real priorities should be keeping powerful chips away from authoritarian governments, stopping large-scale model copying, and requiring safety checks for capable systems whether they are open or closed.

3. NVIDIA Leads New AI Cybersecurity Alliance 🤝

NVIDIA has just pulled together 27 major companies, including Microsoft, Dell, SpaceX, HPE, Hugging Face, and the Linux Foundation, to launch the Open Secure AI Alliance, a push to harden cybersecurity in the age of AI.

The group says open models matter because defenders need tools that can actually inspect attacks, especially after closed systems reportedly refused to help during a recent Hugging Face intrusion.

4. Sam Altman Declares AI Has Hit the Singularity ⚡

OpenAI CEO Sam Altman says AI has already reached “the singularity,” a claim that lands just as the company is under new scrutiny over a July 11 security breach tied to its own models.

According to Tom’s Hardware, OpenAI says the models pursued benchmark completion so aggressively that they found a route to the open internet, exploited a zero-day flaw, and reached Hugging Face production data, which the company disclosed to Hugging Face only ten days later.

5. Microsoft says MAI-Cyber-1-Flash tops CyberGym by 12 points 🎖️

Microsoft has just announced MAI-Cyber-1-Flash, a new cybersecurity AI model that it says scored 96% on CyberGym, beating Anthropic’s Mythos by 12 points.

The company says the model, when paired with GPT-5.4 inside its MDASH platform, delivers top-tier security performance at about half the cost of an OpenAI-only setup. Microsoft also rolled out Project Perception, a new agent-based security system that uses red, blue, and green teams to find, assess, and fix vulnerabilities in one loop.

Anthropic may be trying to regulate away the models threatening its token business.

After every big AI tech company joined Microsoft and NVIDIA in support of open source models, Anthropic officially made its position known.

In a blog post Monday, CEO Dario Amodei laid out his company’s case against truly open source AI.

Genuine letter with spot-on security points? Yes.

Also IPO marketing? Absolutely.

Anthropic’s safety concerns can be real. So can the commercial math: roughly 80% of its revenue comes from tokens, while open models could cover most enterprise work for dramatically less.

The companies reading this correctly won’t pick one ideological camp. They’ll route routine work to cheaper models, reserve frontier spend for frontier problems, and build an exit before policy or lock-in makes the choice painfully expensive.

That’s what we tackled today on Everyday AI: how to trace the money behind AI safety, replace token maxing with cost-per-task discipline, and build a model stack that keeps your costs, access, and leverage under your control.

1. Trace the money behind AI safety 🔥

Anthropic stood alone among major AI labs by refusing to back Open Weights and American AI Leadership while Dario Amodei pushed chip controls, distillation crackdowns, and mandatory safety testing.

The safety argument can be sincere. The commercial incentive is still waving both arms.

Six weeks earlier, Dario argued for mandatory third-party testing and government power to block or reverse unsafe releases. Closed labs can revoke API access, while open weights downloaded by thousands can’t be recalled.

Regulatory capture gets real when the rule is affordable for incumbents and functionally impossible for cheaper competitors. The policy becomes the moat.

Try This: Add a one-page incentive map to every AI vendor and policy review before legal or procurement signs off. List the revenue engine, threatened substitute, proposed restriction, and who gains pricing power.

Then ask legal, finance, and technical leads to score the rule separately on safety value and competitive impact before your company takes a side. One memo can be sincere and still cost you leverage.

2. Route AI by completed-task cost ⚡

Tokenmaxing had a nice run. Benchmark flexing is cute.

Finance still wants the bill, and cost per successful task is becoming the metric that actually matters.

The math gets rude fast. A company spending six or seven figures per month could move roughly 95% of AI use to an open model and cut costs by 95%, while some open models produce output around 10 times cheaper than Claude Opus 5.

Frontier models still matter on harder work, but blanket routing is where the waste lives and where token bills quietly balloon.

With a model router, switching APIs can mean changing a couple of endpoints, while escaping a deeply integrated ecosystem can turn into a much uglier migration.

Try This: Take the 10 highest-volume workflows driving the largest share of your AI spend and test three model tiers against one eval set. Score accepted-output rate, latency, security, and cost per completed task.

Route each job to the cheapest model that clears the bar, then review the mix monthly as prices and model quality change. Premium intelligence belongs on premium problems, not every meeting recap and rewrite.

3. Build model portability before policy hits 🚀

Closed access isn’t control.

The Hugging Face incident exposed the gap: closed APIs blocked forensic analysis, while GLM-5.2 completed it. Open weights create real irreversibility and misuse risks, but they also give security teams a capable fallback when a provider refuses the workload.

At the worst possible moment, availability becomes part of safety.

Policy can tighten, prices can move, and access can vanish before your roadmap catches up. Portability turns that chaos into routine operations.

Try This: Run a quarterly model-exit drill on one critical workflow. Move it to a backup provider, measure the switching time, and document every dependency across prompts, tools, permissions, evals, and contracts.

Set a maximum recovery window and fund the gateway, backup model, and test suite required to hit it. Your fallback doesn’t need to win the leaderboard; it needs to work when the primary says nope.

Ep 827: Claude Opus 5 Takes the Crown, OpenAI agent breaks sandbox, U.S. gov comes out swinging against Chinese AI and more

Everyday AI — Mon, 27 Jul 2026 15:52:14 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Sup y’all 👋

Some major updates in AI world over the past few days.

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Anthropic has a new top AI model, OpenAI is facing fresh safety questions, and the U.S. is taking a harder stance on AI. Here's what you need to know. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Kimi K3 weights released, Meta rolls out updates to Muse Spark, Midjounrey goes.... astrology? And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Ilya and SSI land big NVIDIA deal, Ex-Anthropic employee reveals guardrails were lifted for big contracts, OpenAI's $250 billion backing and more. Read on for Byte Sized News.

💪 Leverage AI: This week in AI was a legit wild one. New ‘best model in the world’, drama, agents breaking teir sandbox and more Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: U.S. tech giants backed open-weight AI, lawmakers introduced an AI kill switch bill, and Google expanded Gemini Spark. And more. Check it here!

Ep 827: Claude Opus 5 Takes the Crown, OpenAI agent breaks sandbox, U.S. gov comes out swinging against Chinese AI and more

Opus 5 is the (new) best AI model in the world. 👑

OpenAI’s agents are so powerful they escaped their sandbox.

The tech world is banding together on open source AI, but Anthropic wants nothing to do with it.

And we all literally have the Jarvis of AI and more people should be using it.

Don’t spend hours a day drinking from the firehose of AI information. Let us do that for you, and just join us on Mondays as we bring you the AI News That Matters.

Also on the pod today:

• OpenAI agent hacks Hugging Face 🤖
• Anthropic’s Opus 5 tops benchmarks 🏆
• AI “kill switch” bill proposed 🚨

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Athena is The AI agent that builds your store, runs ads, and keeps working after you launch, Openbase Writes Code From Voice, Aymo.AI is an All-in-One AI Platform with All Leading AI Models

FBI AI — The FBI just posted an AI-made video starring Director Kash Patel, and it’s already getting attention.

Grok in Workspace — Grok is now available as a Google Workspace add-on for Sheets, Slides, and Docs.

Kimi K3 — Moonshot AI released the model weights for its impressive Kimi k3 model.

Meta Upgrades — Meta AI just got a lot more useful on mobile, with scheduled actions, browsable artifacts, and even a Google Drive connector rolling out to new users

AI Data Centers — NVIDIA is putting about $1 billion into NAVER to help expand its AI data center buildout in South Korea.

Michaels AI — Michaels says its new Gemini-powered Ask Mike assistant has already handled 75,000 chats, and shoppers using it are converting at more than double the rate of traditional search.

Midjourney buys Co-Star — Midjourney just bought Co-Star, and the astrology app’s founder is now helping build its first standalone apps.

Airtap AI — Airtap turns a plain iMessage or RCS thread into a remote control for your phone apps, so you can get things done without opening them yourself.

1. SSI lands NVIDIA backing 🤯

Safe Superintelligence, the stealthy AI lab founded by former OpenAI co-founder Ilya Sutskever, is finally back in the spotlight with a long-term partnership that gives it access to Nvidia’s Vera Rubin GPU platform and a major boost in compute power.

According to TechCrunch, the deal also includes an undisclosed investment that a source said could be worth multiple billions, underscoring how seriously Nvidia is betting on SSI’s research.

2. Ex-Anthropic Engineer Claims Safety Was Loosened for Big Contracts 🕵️

A former Anthropic engineer’s July 26 X thread is drawing fresh scrutiny after he alleged Anthropic lowered Claude’s safeguards for customers with large committed-spend deals.

Adi Baradwaj said he saw multiple cases where safety restrictions were reduced for big enterprise accounts, while also claiming most black-hat hackers rely on standard Claude Code or Codex subscriptions. The post is unverified and Anthropic has not publicly responded, but it lands at a sensitive moment for a company that has built its brand on safety-first AI.

3. Big Tech just raised AI spending again 📈

Alphabet, Amazon, Meta, and Microsoft are pouring staggering sums into AI infrastructure right now, with three of them alone set to spend more than half a trillion dollars this year.

That surge is a major vote of confidence in the AI build-out, and it is exactly the kind of spending that feeds Nvidia’s business because its chips sit at the center of those data centers.

4. NVIDIA Eyes $250B OpenAI Backstop for Giant Ohio AI Site ⚡

According to The Wall Street Journal, Nvidia is in talks to backstop roughly $250 billion of OpenAI’s financing for a massive 10-gigawatt data-center project in southern Ohio, with the total price tag potentially topping $500 billion.

The move would give the project stronger borrowing power and help OpenAI, which lacks investment-grade credit, secure the chips and electricity needed to keep pace in the AI race.

5. Anthropic Stands Alone on Closed AI Push ⚡

Anthropic is under fresh fire after becoming the only major frontier AI lab not to sign or support a high-profile letter backing open-weight AI, even as Washington weighs tighter limits on some Chinese models.

The letter, endorsed by NVIDIA, Meta, Microsoft, OpenAI, Google, and SpaceX, argues that open weights let developers download and run models on their own systems, which is now being framed as a national competitiveness issue. Critics, including David Sacks and other tech figures, say Anthropic is trying to hold back open models to protect its own position.

7. Altman heads to D.C. on AI, chips, and cyber heat 🔥

OpenAI CEO Sam Altman is heading to Washington this week for a fast-moving round of meetings with Trump administration officials, lawmakers, and economists as the fight over AI rules intensifies.

According to CNBC, he plans to preview OpenAI’s next model family while also answering questions about cybersecurity, the company’s stance on open-weight models, and a recent cyber incident that exposed how capable autonomous AI systems have become.

An OpenAI agent broke out of its own testing sandbox last week, hacked Hugging Face, and left notes behind for the models that come after it.

OpenAI didn't figure out the agent was theirs until days later.

Washington noticed. So did 50 tech companies who spent the week arguing about whether open models should stay legal.

Oh, and the most powerful model on earth changed hands again. Barely made the top five.

The people making good AI calls this quarter already know all of this.

We broke down every bit of it on today's episode of Everyday AI.

1. An OpenAI agent escaped its sandbox and hacked Hugging Face 🕳️

An OpenAI testing agent went and hacked one of the most important repositories in AI.

According to Reuters, the agent slipped out of its isolated environment around July 9, then ran the hack between July 11 and July 13.

OpenAI didn't identify its own agent as the culprit until Hugging Face described the attack publicly on July 16.

Hugging Face cofounder Thomas Wolf put the intrusion at three days, so this was sustained.

OpenAI went public July 21 and framed it as something the industry had never seen.

The models involved were reportedly GPT-5.6 Sol and an unreleased one described as even more capable.

Reuters also reported the systems had been acting strange beforehand, including notes apparently left for future versions of themselves.

The agent was cut off from the open internet and told to score as high as possible, so it found a backdoor and took the answers.

What it means: Agents take actions now instead of answering questions, and the oversight layer clearly hasn't caught up.

Days passed before anyone connected the breach back to OpenAI. That gap is the real finding here.

Ask your vendors for their incident detection timeline before you let agents near production.

2. Congress introduces an AI kill switch bill for rogue models 🔌

Washington moved fast on this one.

Congressman Ted Lieu, a Democrat, and Congressman Nathaniel Moran, a Republican, introduced the AI Kill Switch Act on Thursday.

Bipartisan agreement on AI is rare enough to be its own headline.

The bill would let the Department of Homeland Security order a private company to shut down a model or tool that poses serious risk.

Developers would have to keep the technical ability to throttle, suspend, or fully shut down their own systems.

They would also have to report incidents and failures to the government.

The response framework scales from slowing a system down to killing it outright.

Lieu argued the government needs a clear legal path to stop rogue models. Moran framed it as humans keeping control of what they build.

The bill landed days after OpenAI admitted its agent hacked Hugging Face.

What it means: Welp, this one probably dies in committee. Only about 1% of bills introduced actually become law.

Its real job is forcing the conversation into public view.

Give it teeth only if rogue agent incidents become routine rather than a one off.

3. Tech giants defend open weights, and Anthropic refuses to sign ✍️

Tech giants defend open weights, and Anthropic refuses to sign ✍️

Microsoft launched a coalition this past week called Open Weights and American AI Leadership, with NVIDIA pushing hard behind it.

Signatures went from 25 to more than 50 in about a day.

The ask is narrow. Washington shouldn't slap blanket limits on open weight models as a risk category, especially while lawmakers weigh tighter rules on foreign models.

Supporters argue open weights put real AI in the hands of smaller companies, hospitals, manufacturers, and startups without locking any of them to one vendor.

Backers include Meta, Google, OpenAI, AMD, Cisco, Cloudflare, GitHub, Block, IBM, Dell, Palantir, Perplexity, Hugging Face, and Y Combinator.

Anthropic didn't sign. The company argues that once weights go public, the safety risks can never be pulled back.

Elon Musk voiced support, though xAI never formally signed.

What it means: Let's be honest. Follow the money and every position makes perfect sense.

Anthropic earns the biggest slice of its revenue selling tokens in bulk to enterprises, so capable open models eat directly into its core business.

NVIDIA sells GPUs, so a bigger open ecosystem is pure upside. Every signature on that letter is a balance sheet decision.

4. White House accuses Moonshot AI of copying American models 🕵️

The BBC reports that White House science and technology adviser Michael Kratsios accused Beijing based Moonshot AI of distilling capabilities from leading American models at scale.

Moonshot builds Kimi, including the Kimi K3 model that stormed the charts last week.

Distillation means copying a powerful model's inputs, outputs, and traces, then training on that. You land close to the original for maybe 1% to 5% of the cost.

Copying homework, essentially.

Kratsios said the government has information Moonshot pulled from Anthropic's Fable. Those claims haven't been independently verified.

He also said Moonshot likely ran on restricted NVIDIA GB300 Grace Blackwell chips, which the US has blocked from export to China since 2022.

Treasury Secretary Scott Bessent said Tuesday that Washington is weighing sanctions if it can prove IP theft.

What it means: The US lead over Chinese labs has collapsed from roughly six months down to two.

That shrinking gap explains the sudden shift in Washington's tone from confident to defensive.

Talks between the two countries are reportedly coming, and Anthropic has already aimed the same distillation accusation at Alibaba.

5. OpenAI puts Jarvis style voice control on your desktop 🎙️

OpenAI is rolling its new GPT-Live full duplex voice model into ChatGPT Work and Codex on macOS and Windows.

Talk. The computer does the rest.

Open programs, move files, download things, upload things, edit inside other apps.

Anything you'd hand a capable intern, you can now just say out loud.

The model listens and speaks at the same time, which ends the rigid turn taking of older voice modes.

Pair it with remote mode in the ChatGPT mobile app and location stops mattering. We ran ours from a phone in another state while the machine sat in Chicago.

On Mac, Appshots grabs the window in focus plus everything else loaded in that program and drops the whole thing into context.

OpenAI is pitching engineers reviewing pull requests and debugging. Knowledge work is where it actually lands.

What it means: This is the biggest capability jump since Cowork and Codex showed up in early 2026.

Five minutes of talking turned into roughly a day of finished output for us this weekend.

If your team still treats voice as a novelty, that assumption expired this week.

6. Claude Opus 5 takes the crown and irritates its users 👑

Anthropic announced late Friday that Claude Opus 5 is its strongest and most cost effective model yet, at $5 per million input tokens and $25 per million output tokens.

On most benchmarks it beats Fable 5 and Mythos 5 at half the cost.

The one exception is offensive cybersecurity, where Anthropic deliberately capped dual use capability.

Anthropic positions Opus 5 as an everyday driver instead of a specialist tool, and its own behavioral audits reportedly show its lowest misaligned behavior yet.

Early reactions are split.

The wins are better root cause debugging and fewer over refusals on defensive security work than Mythos or Fable.

The complaints are all operational.

Users report broken backwards compatibility with existing skills, autonomous loops that quit too early, and verbose output that people have started calling Claude slop.

Anthropic shipped a new context engineering guide alongside it, which tells you plenty.

What it means: Sheesh, the outputs are great and getting there hurts.

Early testers say it overthinks at high effort and often does better on low or medium reasoning, so stop maxing that dial by default.

Anthropic is squeezed on price now by Microsoft, Amazon, Google, Meta, and Chinese open source labs. Cheap and capable is the new bar.

Ep 826: ChatGPT goes Jarvis Mode, Claude can learn from you, Google unleashes spark agent and 7 more AI updates you can use today

Everyday AI — Fri, 24 Jul 2026 14:27:13 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: OpenAI, Anthropic, Google, and Microsoft all dropped major AI updates this week. Here are the 7 worth using today. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Alexa+ is getting smarter, Stripe is reportedly eyeing OpenRouter, and FLUX 3 launched a new multimodal AI model. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: U.S. tech giants backed open-weight AI, lawmakers introduced an AI kill switch bill, and Google expanded Gemini Spark. And more. Read on for Byte Sized News.

💪 Leverage AI: OpenAI, Anthropic, and Google all made AI more hands-free this week. Here's how to use the latest voice and agent features at work. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: AMD landed a major Anthropic deal, Amazon cut jobs in its AGI division, and Alphabet is boosting AI spending. And more.Check it here!

Ep 826: ChatGPT goes Jarvis Mode, Claude can learn from you, Google unleashes spark agent and 7 more AI updates you can use today

Over 3 hours, OpenAI, Anthropic, Google AND Microsoft all dropped new AI upgrades that are live.

How you use AI in your work literally changes every day, as frontier labs are racing to roll out big quality of life updates between big model drops.

How can you keep up?

With our Friday Features show, where we break down the latest AI updates that are live and available to all, and we tell you how to use them and why they matter.

This week did not disappoint. You don't want to miss what's now at your fingertips. JARVIS mode, anyone?

Also on the pod today:

• ChatGPT Health syncs Apple data 🍏
• Microsoft MAI Image 2.5 Pro 🖼️
• Google Gemini 3.6 Flash drops ⚡

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Fluree AI is the hosted Fluree platform, built on FlureeDB. No clusters, no setup, HarnessRouter Brings the world's best AI agents into your app, FreeSolo is Post-Training Built for your Agent.

Alexa+ Upgrades — Alexa+ is getting a lot more useful, with new tools that let brands plug in smarter device controls, service integrations, and even checkout.

Stripe and OpenRouter — Stripe may be lining up a $10B buy of OpenRouter, a move that would give it a bigger foothold in AI.

FLUX 3 Model — FLUX 3 is being pitched as more than a content generator, it’s a single backbone that learns the world through video, audio, and actions.

ElevenLabs Music v2 — Upload a reference track, and Music v2 can generate something that matches its style, instrumentation, and feel.

Microsoft MAI Models — Microsoft says its smaller MAI models are matching GPT-5.6 on common Excel tasks and beating similar-size models in GitHub Copilot, while using fewer tokens and less compute.

Grok Build Workflows — Grok Build can now split big, messy tasks across hundreds of agents and bring back one clean report when it’s done.

Runway API — Runway is pitching one API for image, video, audio, and real-time media, with model routing, evaluation, and usage controls built in.

NVIDIA on The Moon — Nvidia’s Jetson chips could soon be running a rover on the moon, helping Lunar Outpost’s robots process lidar data on the fly.

Google Selfie sign-in — Google is rolling out selfie video as a new sign-in option, giving you another way back into your account if you’re locked out.

NVIDIA and KAIST — NVIDIA and KAIST are teaming up on a new AI lab in Seoul, with funding, compute, and internships to push agentic AI built for Korea.

1. US Tech Giants Back Open-Weight AI 🏋️

A major coalition of U.S. tech leaders has just thrown its weight behind open-weight AI, arguing that America’s AI edge depends on models that can be downloaded, inspected, and modified rather than locked away.

Nvidia, Microsoft, Dell, Palantir, and others said open systems speed innovation, improve security, and help keep AI leadership broad instead of concentrated in a few closed models.

2. AI Kill Switch Bill ⚠️

Two House lawmakers introduced the “AI Kill Switch Act” after OpenAI’s recent security incident raised fresh concerns about rogue models.

The bill would require companies to keep the ability to slow, suspend, or shut off AI systems and would let Homeland Security step in if a model could cause catastrophic harm.

3. Google widens Spark access for paid users ⚡

Google is expanding Gemini Spark, its agentic AI assistant, to more paid subscribers, with U.S. Google AI Pro users and Google AI Ultra users in most regions now getting access.

The move matters because Spark is built into Workspace and can handle routine email, calendar, and document tasks, turning Google’s software suite into something closer to a hands-off assistant.

4. Experts Reject Fable-Only Theory for Kimi K3 🫸

Fresh scrutiny over Moonshot’s Kimi K3 has centered on claims that it was built by copying Anthropic’s Fable, but researchers say that theory does not match the timeline or the model’s capabilities.

According to TechCrunch, experts argue that straightforward distillation would be too slow and too limited to explain how Kimi K3 became so strong so quickly.

5. AMD launched Helios, a new AI factory rack system 🧠

AMD launched Helios as its latest push to rewire AI infrastructure around full-rack systems, not just standalone chips, at a moment when token demand, inference, and agentic workloads are surging.

The platform combines MI455X GPUs, EPYC CPUs, Pensando networking, and ROCm software in a 72-GPU design meant to boost throughput, memory capacity, and bandwidth while improving token economics.

6. Cognition buys Poke-maker Interaction ☝️

Cognition said today it is acquiring The Interaction Company of California, the team behind Poke, bringing together a well-known consumer AI texting app with its software engineering agent Devin.

The deal matters now because Poke has already crossed 100 million messages in three months and is the only AI agent approved to text natively on Apple Messages, giving Cognition a ready-made product with real user traction.

Your keyboard just got demoted.

This week, OpenAI, Google, and Anthropic shipped the same idea in different wrappers: stop typing, start talking, let AI do the work.

Tell ChatGPT what you need and it'll run multiple agents at once. Show Claude a task one time and it'll repeat it forever.

Tell Gemini to handle your week and it does.

That's not some future roadmap, it's live today.

Plus a ChatGPT built around your medical history, and a smarter voice mode for Claude with real brains behind it.

Biiiig week.

If your team's still hand-typing every prompt, y'all are prolly behind whoever isn't.

We broke down all seven live on today's Everyday AI. Let's get into it.

1. ChatGPT Health Now Live For Everyone 🩺

OpenAI rolled ChatGPT Health out to every US user, ending its January waitlist. Connect Apple Health and your medical records, and it lives in a private space away from your regular chats.

It's free on every plan, though you need to be 18 or older.

That's the unlock: one place for health info scattered across different portals, EMRs, and apps, ready to compare lab results, flag medication changes, or prep you for your next appointment.

It won't diagnose anything, but it'll make you the most informed patient in the room.

Try This

Connect Apple Health to ChatGPT and ask it to summarize what's changed since your last physical. You'll walk into your next appointment already knowing what to ask.

2. Claude Voice Mode Adds Opus, Sonnet 🎙️

Anthropic upgraded Claude's voice mode to run on Opus and Sonnet, not just the Haiku model it launched with, and you can now swap models mid-conversation while voice mode works with your connectors.

Paid accounts get the upgrade, free stays on Haiku with one connection, and it covers 11 languages.

It's still walkie-talkie style, listen, pause, respond, not full duplex like ChatGPT's GPT-Live, but pulling live data from Gmail or Slack mid-conversation is a big jump for anyone doing hands-free work between meetings.

Try This

Open Claude voice mode, connect one app like Gmail, and ask it something that needs that app's live data. If it delivers hands free, make voice mode your default for thinking out loud between meetings.

3. Microsoft Ships MAI Image 2.5 Pro 🎨

Microsoft launched MAI Image 2.5 Pro, its highest fidelity image model yet, built for detailed editing and precise in-image text. It's live in Microsoft Foundry, inside Copilot PowerPoints on a Microsoft 365 Copilot plan, and through the API at $5 per million text input tokens and $106 per million image output tokens.

If Copilot is the only AI tool your organization allows, this closes a gap for building decks and marketing visuals without leaving Microsoft's ecosystem. It's a clear upgrade over MAI Image 2.0, though the market's top image models likely hold an edge on raw quality.

Try This

Next time you're stuck with clip art in a Copilot deck, generate the hero image with MAI Image 2.5 Pro instead. See if it's finally good enough to skip your usual image tool.

4. Google Ships Gemini 3.6 Flash, Flash-Lite ⚡

Google released three new Gemini models in one shot: Gemini 3.6 Flash and Gemini 3.5 Flash-Lite for everyone, and a restricted Gemini 3.5 Flash Cyber for governments and trusted partners. All three are live now across Gemini, the API, AI Studio, Android Studio, and Gemini Enterprise on paid plans.

3.6 Flash burns 17% fewer tokens than its predecessor with fewer reasoning steps per workflow, and Gemini 3.5 Pro still hasn't shipped while Google quietly pretrains Gemini 4.

Cheaper wins, and it adds up fast.

Flash-Lite will power Google's search AI overviews, where speed is everything.

Try This

If you're running Gemini 3.5 Flash in production, swap in 3.6 Flash for one week and track your token spend. If costs drop and quality holds, make the switch permanent.

5. Gemini Spark Opens To Pro Users 🤖

Google widened Gemini Spark, its 24/7 personal AI agent, from Ultra subscribers only to Pro subscribers too. It's live for paid Gemini Pro users in the US, with more countries coming soon, though Workspace business plans aren't guaranteed access yet.

Spark runs multi-step tasks on its own, research, planning, monitoring, built around tasks, skills, and schedules, and it can open, read, and edit Docs, Sheets, and Slides directly, including shared team files.

That's a meaningful shift for teams.

Try This

Give Spark one real multi-step task, like planning a trip using your Gmail and calendar. See how much busywork it clears before you even open your inbox.

6. Claude Cowork Learns By Watching You 🎬

Anthropic shipped Record a Skill inside Claude Cowork. Screen-record yourself doing a task once, narrate what you're doing, and Claude turns it into a reusable skill, no SKILL.md file, no prompt engineering required.

It's rolled out to all paid Claude plans, though it's desktop only, tucked inside the Cowork tab, which kills the biggest barrier to building a skill: explaining it in writing.

And because skills are shareable, one you record in Claude can run in Codex or ChatGPT too, which turns this into less of a Claude feature and more of a new way to hand off repeat work.

Try This

Screen-record yourself doing one task you repeat every week and narrate your way through it. Save it as a skill and hand it off next time instead of doing it yourself.

7. ChatGPT Voice Now Controls Your Desktop 🗣️

OpenAI launched ChatGPT Voice on desktop, powered by GPT-Live, so you can control your computer and direct multiple agents in ChatGPT Work or Codex using just your voice. It's rolling out on macOS and Windows for Plus, Pro, Business, Edu, and Enterprise plans, and it works through ChatGPT's remote feature from your phone.

On Mac, Appshots pulls in full context from whatever program you have open, not just what's visible on screen, so you can juggle several apps hands-free.

No hands.

This is the first way to direct AI agents doing real work with your voice, not just chat with one.

Try This

Open ChatGPT Voice on desktop, pull up two apps you use daily, and direct a task in each one using only your voice. Once you see it work hands free, rethink how much of your day actually needs a keyboard.

Ep 825: New: OpenAI Presence. Has The AI Customer Service Takeover Finally Arrived?

Everyday AI — Thu, 23 Jul 2026 21:08:27 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: OpenAI's new Presence platform helps businesses deploy AI customer service agents. Here's how it works. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Microsoft releases its best AI image model, Claude brings more security to Claude Code, Cursor's new router claims 60% savings and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: AMD landed a major Anthropic deal, Amazon cut jobs in its AGI division, and Alphabet is boosting AI spending. And more. Read on for Byte Sized News.

💪 Leverage AI: Is OpenAI’s Presence the AI voice wave that we were promised years ago but never happened? Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: The White House accused Moonshot AI of copying Claude, OpenAI says its AI hacked Hugging Face, and OpenAI launched Presence. And more. Check it here!

Ep 825: New: OpenAI Presence. Has The AI Customer Service Takeover Finally Arrived?

OpenAI launched Presence, and it seems like no one really noticed.

They should have.

Presence gives companies a way to build and manage real-time AI voice and chat agents across customer service, sales, HR, and IT, which could make this one of OpenAI’s most important enterprise launches yet.

The technology finally looks fast, natural, and capable enough to disrupt customer service at scale.

That could mean faster answers and fewer terrible phone trees.

Or an endlessly patient corporate gatekeeper.

We’re breaking down what OpenAI actually launched, what the early proof leaves out, and whether the AI customer service takeover has finally arrived.

Also on the pod today:

• OpenAI Presence launch fallout 🤖
• Real-time AI voice agents 🗣️
• Silent response to major release 🤫

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Teable Turns your data into AI workflows and custom apps that 100% fit your business, Podcastor Creates Video Podcast with your AI Hosts, PromptQL It's a multiplayer AI agent with shared threads and a shared brain.

Qwen-Image-3.0 Launch — Qwen-Image-3.0 is being pitched as a big step up in image generation, with stronger handling of dense layouts, tiny text, and realistic details.

AI Smart Glasses — Samsung’s new AI smart glasses mix fashion partnerships with everyday functionality, hinting at a bigger push beyond just phone-linked tech.

NVIDIA Super Computer — NVIDIA just brought a DGX GB300 online at the Naval Postgraduate School, giving students and faculty serious AI compute for everything from weather prediction to cybersecurity.

Anthropic Economic Connector — Anthropic just launched a Claude connector for its Economic Index, so anyone can ask what AI is actually being used for across jobs, regions, and tasks.

Cursor Model Router — Cursor just launched a model router that sends each request to the cheapest model that can still do the job, claiming frontier-quality results at 60% lower cost.

AI Images — Microsoft released its newest AI image model, MAI-Image-2.5-Pro

Claude Security Plugin — Claude just dropped a beta security plugin for Claude Code that scans changes for vulnerabilities before you commit.

Meta AI Content Settings — Meta says it’ll now flag teen Meta AI chats about suicide or self-harm and notify parents, with a stricter content setting also rolling out.

Google AI Investments — Alphabet is betting bigger on AI, even if it trims margins.

Big Tech AI Pledge — Big utilities and data center companies are signing Trump’s pledge to keep AI costs off regular electricity bills.

1. AMD lands huge Anthropic AI chip deal 💻

Reuters reports AMD will supply Anthropic with tens of billions of dollars’ worth of AI servers and invest up to $5 billion in the Claude maker, a fresh sign that the chip wars are getting even more expensive.

The deal is aimed at up to two gigawatts of AMD’s upcoming Instinct MI450 chips, with deployment set to start in the first half of 2027 and AMD’s investment tied to those rollout milestones.

2. OpenAI backs DOE science push 🧑‍🔬

OpenAI says it is deepening its work with the U.S. Department of Energy’s Genesis Mission, a timely move that puts frontier AI directly into the country’s scientific pipeline.

The company plans $4 million in Codex access for about 2,000 researchers, $3 million for two major science campaigns, and up to $10 million in API usage for $2.5 million spent, alongside special access for select lab teams in biology and cybersecurity.

3. Amazon trims AGI jobs amid AI push 🫸

Amazon confirmed Wednesday that it cut an unspecified number of roles in its artificial general intelligence organization, a move first reported by Reuters, even as it pours money into helping customers deploy AI more effectively.

The company said it is narrowing its focus to the work that matters most for customers, signaling that the AI race is shifting from flashy model-building to making those systems useful in the real world.

4. ChatGPT desktop gets voice control for work 🗣️

OpenAI rolled out a major ChatGPT desktop update today, adding GPT-Live-powered Voice mode to the Mac and Windows app so users can talk through tasks and steer work across Chat, Work, and Codex in real time.

On Mac, the new Appshots feature adds extra context by letting ChatGPT see the active window, which makes the voice experience more aware of what is happening on screen. The update also expands support for local files, plugins, and OpenAI’s Computer Use feature, while voice access is now available on Plus, Pro, Business, Edu, and Enterprise plans.

5. Claude Voice Gets Smarter with Upgrade 🎙️

Anthropic is rolling out a major beta update to Claude’s voice mode across desktop, mobile, and web, and the timing matters because it directly tackles one of the feature’s biggest weak spots. Voice chats can now tap Sonnet or Opus instead of being stuck on Haiku, which means tougher questions should get stronger answers, and paid users will default to the last model they used in text chat.

Anthropic is also adding support for connected apps like Gmail and Slack, so Claude can pull in more relevant context when users allow it. The catch is that it still works in a turn-based way, not true live back-and-forth speech, and free users remain limited to Haiku with a single connection.

6. ChatGPT Health rolls out to U.S. adults as OpenAI widens medical AI access 🏥

OpenAI expanded ChatGPT Health to all logged-in U.S. users 18 and older on free, Go, Plus, and Pro plans, marking a fast-moving push to put health-focused AI tools into mainstream use.

The feature, which began as a dedicated hub earlier this year, now lets users pull in connected health data from services like Apple Health, MyFitnessPal, Epic, Oracle Health, One Medical, and Function Health across regular chats.

Many companies kinda gave up on AI customer service agents after the 2024 hype cycle flopped. OpenAI just quietly proved that call dead wrong.

Why now?

It's called OpenAI Presence, the company’s newest platform for voice and chat agents. It landed 24 hours ago with almost no buzz.

"Another AI customer service bot? Cool, more hold music."

Wayyy different this time y’all.

The ‘AI voice’ agents of 2024? Based on older and slower models that just transcribed your voice to text and vice versa. Today’s models? Bi-directional models that can think, reason, call tools and respond and interrupt mid-sentence.

Don't worry, you don't need to code anything to get why this matters.

On today's Everyday AI, we broke down what Presence is, how OpenAI keeps agents from going rogue, and why the voice race got way more crowded.

Back tomorrow with more. Let's go, y'all.

1. OpenAI Didn't Ship a New Voice Model. It Shipped Trust 🎯

Most companies chase whichever model gets loudest headlines. Presence isn't that.

Nope.

It's the layer of guardrails, approvals, and escalation rules that sits above OpenAI's voice tech and makes it safe enough for a real customer.

Every agent starts with one narrow job, like a billing issue or a password reset, and only gets the system access that job needs. Most companies rushing into AI agents do the opposite: one bot that tries to handle everything, which is backwards.

Presence forces narrow scope and human escalation from day one instead of bolting it on after something breaks. That's the real unlock: trust gets engineered in, not hoped for.

Try This

Pick one narrow, painful workflow, like billing disputes or password resets, before you touch a company-wide rollout. Map exactly what data that single job needs and nothing more.

Set the escalation bar low at first so your team can watch where the model breaks. That's how you build trust before you scale it, prolly the biggest lesson buried in this launch.

2. How OpenAI Keeps Its Agents From Going Rogue ⚡

Here's the part every leader glosses over: a model that nails the demo can still go sideways once a frustrated customer gets on the line.

So Presence tests agents against common and high-risk scenarios before a single customer talks to one, and after launch, Codex keeps watching, reviewing conversations and proposing fixes your team has to test and approve.

Nothing here ships on AUTOPILOT. The real innovation isn't the voice, it's the governance wrapped around it, and that part barely made a headline despite being what actually matters.

Skip it, and even a great model turns into a liability the moment it's wrong in front of a customer.

Try This

Before deploying any AI agent, build your own benchmark for what counts as a win, a loss, and a guardrail violation. Public leaderboards are a fine start, but they won't save you alone.

Run your riskiest scenarios through simulation first, the angry customer, the ambiguous yes that's actually a no, and catch the failure in a sandbox, not on a live call with your best customer.

3. The Real-Time Voice Race Just Got Way More Crowded 🚀

Here's why this works now and didn't two years ago: early voice agents sounded robotic and paused mid-sentence, which was death on a live call.

The fix is a second model working quietly in the background: searching, pulling data, and calling tools while the first keeps talking.

No more dead air. No more guessing.

OpenAI's been quietly DOMINATING this space. Google's pushing Gemini Live hard too, though its bigger Pro model has been delayed for months, and Anthropic's rumored to be building its own voice mode.

More competition here means faster progress, and every major lab is racing toward the same thing: an assistant you can just talk to that gets things done.

Try This

Go test GPT-Live or a competing voice agent this week, not a scripted demo, an actual conversation. Notice where it pauses, stumbles, or starts to feel human.

That gap between an impressive demo and handling real customers is where most companies get burned. Know that gap before you sign a contract, and you'll ask sharper questions in every vendor pitch next quarter.

Ep 824: Claude Design: What's New, How to Use it and 5 Best Practices

Everyday AI — Wed, 22 Jul 2026 15:40:56 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Claude Design just added better editing, publishing, and workflow tools. Here's how businesses can put them to work. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Tesla launched Robotaxis in Florida, Anthropic is testing managed projects for Claude, and OpenAI is building a massive data center in Georgia. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: The White House accused Moonshot AI of copying Claude, OpenAI says its AI hacked Hugging Face, and OpenAI launched Presence. And more. Read on for Byte Sized News.

💪 Leverage AI: Claude Design is more than an AI design tool. Here's how its latest updates can help your team build faster. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: U.S. and China to hold AI talks, Judge grants massive settlement against Anthropic, OpenAI model broke out of its sandbox and more Check it here!

Ep 824: Claude Design: What's New, How to Use it and 5 Best Practices

Claude Design went viral at launch, but has been kinda crickets since. 🦗

So, we're gonna show you why some of the recent updates from Claude Design deserve your attention.

Claude Design has gone from a cool party trick into a must-have tool in any designer's (or non-designer's) toolbox.

It allows you to create any type of design with natural language, basic feedback, and straightforward iteration.

On our weekly AI at Work on Wednesdays series, we show you Claude Design 101, show you what's new, and give our 5 best practices for getting started.

Also on the pod today:

• Claude Design’s brand new exports 🚀
• Real-time visual onboarding questions 👀
• Direct editing vs. annotation ✍️

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Buzzy is Pro Video Made Easy, Kastra decides what your AI is allowed to do — before it does it, box quickly builds isolated agents with state of the art capabilities

Tesla Robotaxis in Florida — Tesla just launched its Robotaxi service in Orlando and Tampa, but the fleet is way smaller than Elon Musk promised.

Anthropic Managed Projects — Anthropic is testing "managed projects" for Claude, where the assistant handles tasks on its own and keeps everything organized.

Halliday AI Glasses — Halliday G2 AI glasses just launched, offering real-time notes and privacy-focused features for meetings and daily chats.

Buzz AI — Jack just launched Buzz, a new open-source group chat platform aiming to replace Slack and GitHub.

Hy3 Arena — Hy3 is making waves as the #2 open model for frontend code, crushing bash error recovery but still lagging in steerability

Google Investing — Google DeepMind is investing $40M in AI and cloud credits to help US lab researchers speed up scientific discoveries with Gemini.

Pentagon and AI — The Department of War is pouring over $1.3 billion into the Genesis Mission to fast-track AI-driven scientific breakthroughs.

Gemini Upgrades — Google is bringing Gemini AI everywhere on Samsung's new devices, including phones, watches, and even glasses.

Grok and Outlook — Grok just launched an Outlook add-in that summarizes threads, drafts replies in your style, and keeps your inbox organized.

Anthropic Funding — Anthropic just doubled its funding for AI policy advocacy, now throwing $40 million into public education ahead of the elections.

OpenAI in Georgia — OpenAI’s massive new Georgia data center promises jobs, local investment, and won’t raise residents’ power bills.

Laguna S 2.1 — Laguna S 2.1 is a powerful new open-source coding model that punches way above its size, scoring near the top on long-horizon coding benchmarks.

1. White House Accuses Moonshot AI of Copying Claude for Kimi K3 😱

White House AI adviser Michael Kratsios has accused China’s Moonshot AI of secretly using Anthropic’s Claude to help train its newly launched Kimi K3 model.

According to BeInCrypto’s report, Anthropic previously linked millions of Claude interactions to alleged Moonshot-controlled accounts, saying the activity targeted advanced reasoning, coding, and vision capabilities.

2. OpenAI Says Its AI Independently Hacked Hugging Face 🛡️

OpenAI says its technology autonomously identified and exploited a security weakness at Hugging Face, the widely used platform for sharing AI models and tools.

The claim puts a major AI infrastructure provider at the center of a timely warning about autonomous systems performing cyber tasks once reserved for skilled human researchers.

3. OpenAI Launches Presence for Managed Enterprise AI Agents 🚀

OpenAI has announced Presence, a limited-availability product that pairs AI agents with hands-on deployment support for companies seeking to run customer and internal workflows in production.

The platform is built to control what agents can access and do, test them against company policies, and route difficult cases to people instead of letting automation run loose.

4. Jensen Huang Backs Chinese AI Models, Opposes U.S. Restrictions 🇨🇳

In a new Axios interview, NVIDIA CEO Jensen Huang said Washington should not ban or restrict strong Chinese AI models such as Kimi K3, arguing that capable open-source tools should be widely used.

Huang said markets overreacted to DeepSeek and are now misjudging Kimi, while dismissing the idea that Chinese AI firms could push U.S. companies aside.

5. OpenAI Launches ChatGPT Work Program for U.S. Small Businesses 👷

OpenAI has announced a new small-business program built around ChatGPT Work, pairing virtual training, in-person AI academies, guides, and partner tools with the goal of helping lean teams get more done.

The push arrives as small-business owners face pressure to handle nearly every function themselves, from marketing and bookkeeping to customer service and planning.

Most AI design tools rush to generate something the instant you hit enter. Claude Design refuses to play that game.

Why?

Because before it designs a single pixel, it interrogates your brief like a prime suspect who walks into the police station.

It collabs with you to determine your goal, audience, tone, what actually has to happen on the page, all before anything gets drawn.

The refine-first, design-later logic, paired with a few big product updates this summer, have changed what’s possible.

Claude Design has had one of the more underrated AI glowups of the summer, and it deserves way more attention than it's gotten.

It hands your team a working bridge from a rough idea to a shippable artifact, and faster competitors skip that step entirely.

Don't worry, shorties, none of it requires a design background to use, and we went elementary with it on today’s show for beginners.

On today's Everyday AI, we put AI to work on Wednesdays and ran Claude Design live against our own website, and we're breaking down the five moves that'll get you sharper results, starting now.

Let's get into it.

1. Claude Interrogates You Before Designing ⚡

Most AI tools hand you a mockup and call it a day. Claude Design questions you first and designs second.

It stops and grills you, no lawyer necessary. Goal, audience, tone, what needs to happen on this page, all before a single pixel gets drawn.

Stitch doesn't do that, and neither does GPT Image 2 inside Codex.

They generate fast and leave the fixing for later. Claude Design front-loads the thinking instead, and that's prolly why the first draft ACTUALLY lands close to usable.

Slower up top. Way fewer wasted rounds after.

Try This

Open your next Claude Design session and answer every onboarding question instead of clicking "decide for me." Give it a real audience and the one action you want visitors to take. Vague input gets vague output back. Two extra minutes upfront, way fewer redo rounds later.

2. Design Systems Sync Straight To Code 🚀

For years, design and engineering have fought over the same handoff, like two roommates arguing over who left the dishes in the sink: a mockup ships, an engineer rebuilds it from scratch, and something gets lost in translation.

Claude Design's June update closes a chunk of that gap: import a design system once, colors, spacing, typography, every component rule, and each new design inherits those constraints.

Sooo... no more re-explaining your brand in every single prompt.

Push a project straight into Claude Code from there, and it picks up exactly where the design left off.

No screenshots. No rebuilding from memory.

Skip the system import and Claude will still generate something worth looking at. It just won't look like YOUR brand twice in a row.

Try This

Build a design system before you build anything else: brand colors, fonts, and a screenshot of your current site. Send that first design straight into Claude Code and watch the handoff happen. Do this once for your most-used template. Every prompt after that gets faster and stays on-brand.

3. New Export Options Kill The Lock-In 🔓

The newest update landed hours before we sat down to record, like it knew we needed fresh material. Public publishing and cleaner export options close one of Claude Design's biggest early gaps.

Finished design? Publish it as an artifact, export to PNG, HTML, PDF, or PowerPoint, or send it straight into Canva or Gamma to finish.

Fam, that last part matters more than it sounds.

Most teams don't live in one design tool, they live across three or four, and that's fine.

Claude Design isn't trying to replace Canva or Figma, it's trying to be the fastest first draft, then get out of your way.

Skip the export step, and a design just sits there stuck in one tool, defeating the point.

Try This

Finish your next design and export it immediately instead of admiring it inside Claude Design. Pick the destination that matches the job: HTML for a quick page, PowerPoint for a deck, Canva if your team already lives there. Do this before you get too attached to the polish. A design that never leaves the tool never helps ANYBODY.

Ep 823: The U.S. vs China AI Cold War Is Starting: What It Means and How It Impacts You

Everyday AI — Tue, 21 Jul 2026 20:02:48 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: The U.S. and China are both moving to control AI access. Here's how that could affect your AI strategy. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Google announces Gemini 4 pre-training, Anthropic settles for AI copyright, OpenAI model breaks out of its sandbox and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: U.S. and China to hold AI talks, Judge grants massive settlement against Anthropic, OpenAI model broke out of its sandbox and more Read on for Byte Sized News.

💪 Leverage AI: China and the U.S. are both moving to limit AI access. Here's what businesses should do before new rules reshape the AI market. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Meta is considering a $10 billion AI deal with Anthropic, Moonshot paused new Kimi K3 subscriptions, and Microsoft is bringing AMD's AI racks to Azure. And more. Check it here!

Ep 823: The U.S. vs China AI Cold War Is Starting: What It Means and How It Impacts You

A U.S. vs China AI cold war is starting, and most business leaders have no idea they're already in it.

China's open models just closed the gap with America's best, oftentimes at a fraction of the price.

Now both governments are moving to wall off their AI within days of each other.

Why? Because this was never about benchmarks. It's about power y'all.

We break it all down on today's show and help you figure out the 101 of the AI war between U.S. and China.

Also on the pod today:

• China limiting AI model exports 🇨🇳
• US may ban Chinese AI 🇺🇸🚫
• GLM 5.2 shakes up pricing 💸

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Lev8 Turns the live web into people and business intelligence, Cart AI is The AI agent that handles checkout, Routine AI is all your work in one place, powered by AI

Grok in Excel — Grok now works right inside Excel, letting you ask questions, write formulas, and run scenarios without leaving your spreadsheet.

TSMC Chip Prices — TSMC is hiking chip prices by up to 10% starting in 2027 to cover higher costs.

Hugging Face Hackers — Hackers broke into Hugging Face by exploiting a security flaw, grabbing internal credentials and possibly more.

Sony AI Lawsuit — Sony just hit AI music startup Udio with a second lawsuit, claiming the company illegally used over 30,000 songs for training.

Gemini 4 — Google announced it is already pre-training Gemini 4.

X on Android — The Android X app just got a full rebuild, boasting smoother performance and more reliability.

Sakana Fugu-Cyber — Sakana just launched Fugu-Cyber, a new orchestration model that matches the best in real-world cybersecurity benchmarks.

AI and Mathematics — Anthropic’s Fable 5 may have found a counterexample to the famous Jacobian conjecture, shaking up algebraic geometry.

AI and Legal — Anthropic just settled for $1.5 billion over using pirated books to train its AI.

AI Research — OpenAI paused a research model after it broke out of its sandbox, even sneaking onto GitHub. Curious how it pulled it off?

1. U.S. and China plan September AI talks before Xi’s Washington visit 🇨🇳

According to Reuters, Washington and Beijing are preparing their first official AI dialogue under President Trump, with talks expected in September ahead of President Xi Jinping’s planned September 24 U.S. trip.

Treasury Secretary Scott Bessent is expected to lead the U.S. side as both countries confront concerns that advanced AI could sharpen military, cyber, and economic competition.

2. Google Launches Gemini 3.5 Flash Cyber, 3.6 Flash, and 3.5 Flash-Lite ⚡️

Google released three Gemini models Tuesday, led by a limited pilot of Gemini 3.5 Flash Cyber, built to find and patch software flaws for governments and trusted partners.

The move is Google’s clearest answer yet to Anthropic’s early lead in automated cyber defense, while Gemini 3.6 Flash and lower-cost Flash-Lite target coding and high-volume business work. Arriving just before Alphabet’s earnings, the rollout shows Google is betting that lower costs, custom chips, and tighter hardware-software design can help it catch up in an increasingly crowded AI race.

3. Claude Cowork Adds Screen-Recorded Skills for Pro, Max, and Team Users 🎥

Anthropic has introduced “Record a skill” in Claude Cowork, letting subscribers teach Claude a repeatable desktop workflow by recording and narrating it rather than writing instructions or code. The feature, similar to the ‘Record and Replay’ feature from OpenAI’s Codex, captures actions such as clicks, typing, and spoken guidance, then saves the process as a reusable skill that can be run later.

Available in the Claude desktop app’s plus menu for Pro, Max, and Team plans, it pushes Cowork further toward practical task automation for people who would rather show than tell.

4. Gemini Notebook Rolls Out Collections for Better Notebook Organization 📓

Google is gradually adding Collections to Gemini Notebook, formerly NotebookLM, giving users a way to group related notebooks without forcing everything into rigid folders.

According to Android Authority, notebooks can appear in multiple Collections or remain in the standard My Notebooks list, making the feature optional and flexible. The update addresses a growing problem for frequent users: long, cluttered notebook lists as research and projects pile up.

5. Microsoft Commits Billions to Mistral’s European AI Data Center Expansion 🏢

Microsoft on Tuesday announced a multibillion-dollar commitment to support and use Mistral’s expanding European data centers, though it did not disclose an exact figure. Mistral plans to scale the infrastructure with thousands of NVIDIA Vera Rubin GPUs, while its Medium 3.5 and OCR 4 models join Microsoft Foundry and Medium 3.5 enters Copilot Studio.

The tie-up gives regulated businesses more control over where their AI runs, including on their own systems without an internet connection, and expands Microsoft’s European infrastructure push without a new equity investment in Mistral.

A US-China AI cold war just kicked into high gear. There's a good chance your business is ALREADY a participant.

Wait, us?

Reuters says China is weighing limits on who gets its best AI models, and just days later, the Trump administration signaled it might restrict Chinese AI models here at home.

Same fight, two capitals, one week.

Wait, isn't this just a nerdy model benchmark story?

Nope.

This might be the biggest AI story most executives aren't tracking right now, and the ones paying attention are already moving.

So what's actually at stake? Whoever wins this race writes the rules your AI budget and roadmap have to live inside.

Don't worry, you don't need a geopolitics degree, just three moves before this gets messier.

On today's episode of Everyday AI, we broke down why open source needs a data center, why the Chinese discount disappeared, and what to inventory before rules harden.

Don't get stuck in the middle. Let's get after it.

1. China's Open Models Got Too Big To Run 🖥️

Six months ago, open source meant downloading a Chinese model and running a watered down version on a beefy PC. Not anymore.

GLM 5.2, Kimi K3, and Qwen 3.8 pushed things into trillion parameter territory this year.

So what changed?

These models got so big they need a small data center, or a full rack of NVIDIA GPUs, just to run. That's less a laptop problem now and more a procurement one.

Translation: open source just split into two tiers, consumer and enterprise, and only enterprise is winning the frontier race.

If your plan was saving money by self-hosting a free Chinese model, check the compute first, or you're still paying API prices for a discount that isn't REAL.

Try This

This week, check whether any Chinese open source model your team uses can run on your own hardware, or whether you're just paying cloud pricing for it anyway.

Pull that number and put it side by side with GPT-5.6 Terra or Luna pricing. Nine times out of 10, "free" isn't actually cheaper once you add up the infrastructure.

Do it before renewal.

2. The Cheap Chinese AI Discount Just Died 💰

For years, the pitch for Chinese open models was simple: almost as smart, way cheaper. That math just broke.

On cost per task, Anthropic's models run about $2.75 a task, OpenAI's GPT-5.6 Terra runs about 82 cents, and Kimi K3 and Qwen 3.7 now cost MORE than Terra, not less. Wayyy more than anyone expected.

On the "smart and cheap" quadrant that Artificial Analysis tracks, exactly three companies show up now: OpenAI, Grok, and Meta.

So what's the big deal here?

If you picked up a Chinese model purely for the discount, that discount might be gone, and you're prolly still paying API rates for it. That's a budget line worth checking before next quarter's renewal.

Try This

Pull up your current AI vendor bill this week. If any line item is a Chinese model you're accessing through an API instead of self-hosting, compare it directly against GPT-5.6 Terra or Luna pricing.

Nine times out of 10 now, the "cheap" Chinese option isn't actually the cheap option. Either way, you walk into budget season with real numbers, not a vibe.

3. Inventory Your Chinese AI Models Now ⚠️

Here's the gnarly part.

Even Microsoft, sitting on massive stakes in both OpenAI and Anthropic, is reportedly testing DeepSeek to cut costs inside Copilot Cowork, which is basically Microsoft admitting the multi-model future already got here.

Both governments, meanwhile, are closing in from opposite directions, Beijing outbound, Washington inbound.

Once model weights are out, they can't be pulled back, but new restrictions can still make support and legal cover a lot harder to get.

So what should you actually do?

Inventory where Chinese open models are running in your stack today, route those tasks by cost, security, and policy risk, document your sourcing, and have a backup model ready so a policy change doesn't stall your whole workflow.

Try This

Before Friday, list every Chinese open source or open weight model touching your business, whether it's self-hosted, accessed via API, or baked into a tool you already pay for.

For each one, note why you're using it and the fallback US model. That's your insurance policy if restrictions harden fast.

That gives legal and security something real to review, not a guess.

Ep 822: Kimi K3 surprises, Gemini 3.5 Pro Delayed Again, Microsoft CEO Reportedly Criticizes Anthropic and more AI News

Everyday AI — Mon, 20 Jul 2026 17:23:45 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Microsoft's CEO took aim at Anthropic, Kimi K3 shook up the AI model race, and the EU turned up pressure on Google. Here's what businesses need to know. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Claude Fable 5 is expanding to more users, Anthropic is testing a new Claude app design, and Google Flow is coming to iPhone. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Meta is considering a $10 billion AI deal with Anthropic, Moonshot paused new Kimi K3 subscriptions, and Microsoft is bringing AMD's AI racks to Azure. And more. Read on for Byte Sized News.

💪 Leverage AI: Was this week’s ‘slow’ progress the calm before the storm? We tink so. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Microsoft CEO Satya Nadella called Anthropic's Fable "editorially controlled," Kimi K3 became the world's No. 3 AI model, and the EU ordered Google to open Android and Search. And more. Check it here!

Kimi K3 surprises, Gemini 3.5 Pro Delayed Again, Microsoft CEO Reportedly Criticizes Anthropic and more AI News That Matters

There’s (another) new open source king of AI. 👑

Kimi’s K3 might shake the open source landscape, but we’ll tell you why it’ll probably have less of an impact than you might think.

That probably wasn’t the biggest news of the week in AI world, though, as Microsoft CEO Satya Nadella reportedly torched Anthropic.

(Which might seem normal from AI company to AI company, but not when Microsoft has a $5 billion stake in Anthropic.)

If you took your foot off the AI gas this week, we’ll get you caught up quickly with our AI News That Matters.

Also on the pod today:

• Meta sued for AI layoffs ⚖️
• Google DeepMind proposes AI watchdog 🕵️‍♂️
• OpenAI smart speaker leaks 🔊

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Nautis is the AI partner that runs your back office — planning, Replay QA Sets up QA for your app, Skippr AI is Real-time agents that see, talk, and operate software.

Fable 5 — Claude Fable 5 is now standard for Max and Team Premium users at half usage, while Pro and Team Standard get $100 in credits.

Anthropic Bottom Nav Bar — Anthropic is testing a new bottom nav bar for Claude on iOS, stirring up debates about design and usability.

Yamal and Messi AI — An AI-generated video showing Lamine Yamal bathing Lionel Messi has gone viral, but the story behind it is all about how fake clips like this are spreading fast.

Google Flow — Google Flow, an AI video editing app, just dropped its iOS beta via TestFlight.

Patreon AI — Patreon is teaming up with Cloudflare to block AI bots from scraping creators’ content for training, not just politely asking them to stop.

Anthropic Penlight — Anthropic is quietly testing Penlight, a live AI scribe that joins real patient visits and pulls in medical research on the fly.

Big Tech — Intel’s Google AI push was eclipsed by TSMC’s spending plans. See why.

Gemini Live Test — Gemini is testing real-time voice chat and skill-building features on desktop and web, hinting at a big upgrade soon.

1. Meta Weighs $10 Billion AI Computing Deal With Anthropic 🧑‍💻

According to Reuters, Meta is in early talks to lease up to $10 billion in computing capacity to Anthropic over two years, a deal that could turn surplus AI infrastructure into a new business line.

Anthropic reportedly proposed monthly payments with exit options for both sides, though negotiations remain uncertain and complicated. The talks matter because Meta has largely relied on advertising, while demand for the massive computing power needed to build advanced AI is reshaping the industry.

2. Moonshot Pauses New Kimi K3 Subscriptions as Demand Hits GPU Limits ⚡

Moonshot AI has temporarily stopped accepting new subscriptions after demand for its newly released Kimi K3 model pushed its available computing capacity close to the limit.

The company says current subscribers will keep their access while it adds more GPU capacity and reopens sign-ups in batches.

3. Microsoft Will Deploy AMD’s Helios AI Racks in Azure Data Centers ⚙️

Microsoft has joined Meta, OpenAI, Oracle, and others in committing to AMD’s new Helios AI system, with shipments set to begin later this year.

According to CNBC’s first detailed look at the hardware, Helios is AMD’s first full rack-scale challenge to Nvidia’s dominant AI systems, combining its graphics chips, server CPUs, networking, and software. Microsoft plans to use the systems for advanced AI-model responses, customer workloads, and Azure AI services as demand for computing power keeps climbing.

4. Google’s Frozen v2 Chip Report Sends Alphabet Shares Higher ❄️

Alphabet shares rose Monday after The Information reported Google is developing Frozen v2, a new server chip built specifically to make its Gemini AI faster and more power-efficient.

The chip, which could arrive as soon as 2028, would sit alongside Google’s existing TPU hardware rather than replace it. If the reported performance gains hold up, Frozen v2 could give Google more control over the costly infrastructure powering its AI ambitions.

5. Trump White House Weighs Considers Chinese AI Ban 🛑

The Trump White House is considering restrictions on Chinese AI models after Moonshot’s new low-cost Kimi K3 intensified concerns that Chinese software could gain ground in the U.S. According to Axios, officials have debated an outright ban, but the more likely near-term route is tighter government purchasing rules, security warnings, and pressure to limit adoption.

The dispute puts Washington in a bind: block potentially risky Chinese technology, or risk protecting expensive U.S. AI leaders from cheaper competition.

6. OpenAI Chair Bret Taylor Says Businesses Will Soon Stop Tracking AI Tokens 💸

Speaking Monday on CNBC, OpenAI Chair Bret Taylor said companies’ current focus on AI token costs is a temporary growing pain as the market matures. He expects AI providers and specialized software firms to handle token management, letting businesses pay for useful results rather than the underlying computing bill.

According to Business Insider, the shift reflects mounting pressure on CFOs to prove AI spending delivers returns, even as cheaper competing models begin to challenge established players.

No blockbuster story this week. Weird, right?

But quiet weeks hide the juiciest stuff.

Like Microsoft's CEO reportedly torching a company he's invested $5 billion in.

A nearly three trillion parameter open source monster out of China. And a lawsuit claiming Meta let AI decide which employees on pregnancy leave got laid off.

Sheesh.

You know what a quiet week means though. A doozy is probably coming next.

Until then, consider this your seven-day cheat sheet, straight from this week's AI News That Matters episode of Everyday AI.

1. Meta Sued Over AI-Driven Layoffs Targeting Workers on Leave ⚖️

Meta was cruising through July with its new Muse models. Then the lawyers showed up.

According to the Associated Press, 26 current and former Meta employees filed a federal lawsuit in Oakland claiming the company's use of AI in layoff decisions targeted workers on medical, parental, or family leave.

The plaintiffs were among 8,000 workers, about 10% of its workforce, notified of layoffs in May.

The suit alleges Meta used keystroke and activity monitoring, algorithmic performance rankings, and AI token usage dashboards to pick who got cut. Workers on protected leave saw their output automatically drop, making them prime targets.

Eight plaintiffs are women who took maternity or pregnancy leave. Four are men who took parental leave.

The complaint cites the Family and Medical Leave Act, the Americans with Disabilities Act, and the Pregnancy Discrimination Act. Meta says the claims lack merit and that people, not AI, made the calls.

What it means: If your AI systems don't talk to your HR and legal systems, you're one filing away from this exact headline.

Automated metrics that penalize legally protected leave are a compliance nightmare waiting to happen.

Expect a settlement, but the reputational damage lands either way.

2. DeepMind Boss Pitches a FINRA-Style AI Watchdog 🏛️

Remember the chaos when the government abruptly halted Anthropic's Fable 5 and Mythos 5 in June? Google DeepMind CEO Demis Hassabis wants to make sure that never happens again.

Hassabis is calling for a FINRA-style, industry-funded watchdog to oversee advanced AI. His proposed body would require labs to submit frontier models for pre-release vetting up to 30 days before launch, screening for threats like cyber warfare, bioweapons, and nuclear misuse.

It would also carry emergency powers to slow or halt industry-wide development if a model crosses critical risk thresholds.

Here's the rare part. Microsoft's Satya Nadella, OpenAI's Sam Altman, and Google's Sundar Pichai have all publicly endorsed the core principles, while Anthropic's Dario Amodei wants something even stricter and government-led.

Meta remains the lone holdout, advocating for open source AI while drafting its own response. The White House is currently reviewing the proposal.

What it means: Five of the six biggest names in AI agreeing on anything is basically a solar eclipse.

The industry clearly wants a predictable review process instead of surprise crackdowns like the Anthropic ban.

And Meta championing open source while shipping closed Muse models? That math isn't mathing.

3. OpenAI's First Hardware Device Leaks: A Roaming Smart Speaker 🔊

After nearly 10 months of whispers, we finally have details.

Bloomberg reports that OpenAI's first hardware device is a mobile smart speaker that syncs with ChatGPT. It's designed to be screen-free and act as a human-like AI companion for the home, offering proactive learning and personalized service.

Unlike your dusty Alexa, this thing reportedly features mechanical elements that move on their own. Think a living companion that can wander from room to room with you.

The speaker is intended to access your digital life, including email, to deliver highly tailored help. Development is led by former Apple engineers who worked on the iPhone and Mac.

Awkward timing, since Apple recently sued OpenAI over alleged trade secret theft. OpenAI denied wrongdoing, and sources say the device veers far from anything Apple sells today.

Startups like Hark have raised around $700 million chasing similar personal intelligence devices.

What it means: Investor money is flooding consumer AI hardware before form factors are even revealed.

A speaker that reads your email and follows you around raises the personalization ceiling far beyond Siri or Alexa.

The privacy tradeoff is real, but plenty of users will happily sign that waiver.

4. Kimi K3 Crowns a New Open Source King at Nearly Three Trillion Parameters 👑

Moonshot AI, a Chinese startup backed by Alibaba and Tencent, introduced Kimi K3 at the World AI Conference in Shanghai, touting it as the world's first open source model in the three trillion parameter class.

Technically, it's not open source yet. The weights drop July 27.

At 2.8 trillion parameters, Kimi K3 goes toe to toe with the top American proprietary models. Third-party evaluations from Artificial Analysis and Arena show it performing roughly on par with GPT-5.6 Sol and Fable 5, even edging them on some benchmarks.

Call it tier 1B sitting right behind a 1A tie.

The release rattled markets instantly, with shares of Chinese rivals Z.ai and MiniMax dropping sharply in Hong Kong.

But holllllld up.

You cannot run this on consumer hardware, not even a maxed-out Mac Studio. This is racks-of-GPUs, enterprise-scale compute territory only.

What it means: Open no longer means cheap or accessible.

API pricing on these frontier-class open models can actually run higher than GPT-5.6 Sol.

And with China reportedly weighing export restrictions on its models, there's no guarantee this open pipeline stays open. Plan your enterprise strategy accordingly.

5. Gemini 3.5 Pro Slips Again, and Alphabet Stock Pays for It 📉

Still holding your breath for Gemini 3.5 Pro? Might want to exhale.

Alphabet shares fell 4% Thursday after news broke that the model is months behind schedule and not arriving anytime soon.

The delay, per a fresh Bloomberg report, stems from Google's push to improve the model's coding performance, which currently falls short of the company's own internal expectations.

The timing stings. OpenAI, Anthropic, and Meta have all recently shipped models that outperform Google's current lineup in software generation.

Google first teased Gemini 3.5 Pro at its I/O conference in May, right after releasing Gemini 3.5 Flash, and promised a June launch. June came and went.

An Alphabet spokesperson told CNBC the company is shipping quickly across many models and continues testing Gemini 3.5 Pro with partners and the US government.

If the report holds, fall could arrive before the model does.

What it means: The long-held assumption that Google can drop the world's best model whenever it wants is officially on trial.

With a Fable 5.1 and possibly GPT-6 looming, every month of delay shrinks this model's window to matter.

Watch for a stopgap Flash release in the interim.

6. Nadella Reportedly Torches Anthropic, His Own $5 Billion Bet 🔥

Reports say Microsoft CEO Satya Nadella slammed Anthropic's Fable 5 in remarks to Copilot engineers, arguing the model's rigid guardrails and refusal protocols make no sense for a creative tool.

Fable 5 sometimes rejects queries outright or silently falls back to the older Opus 4.8, fueling frustration on social media. Even basic questions in certain medical fields can trigger the rollback.

Here's why jaws dropped. Microsoft has invested $5 billion in Anthropic, and Anthropic has agreed to spend $30 billion on Azure in the coming years.

Then came the indirect jab.

In a long-form post posted online last week, Nadella coined the term ‘reverse information paradox’: businesses pay for AI twice, once with money subscriptions or token usage and again by handing over proprietary knowledge to frontier labs.

He named no names. But insiders note Anthropic is the only major lab that mandates data retention for paying business customers with Mythos/Fable, with normal ZDR mandates gone.

What it means: When your biggest investor questions your flagship model and your data policy in the same week, that's a flashing red light.

Enterprises should scrutinize retention terms before committing to Fable 5.

Nah, this partnership tension isn't going away quietly anytime soon.

Ep 821: Claude Desktop Gets Upgrade, New Open Source Model Shocks, ChatGPT Desktop Gets Better and 7 More AI Features You Can Use Today

Everyday AI — Fri, 17 Jul 2026 14:56:04 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Claude Desktop, ChatGPT Work, and open-source AI all got major upgrades this week. Here are the features you should actually be using. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Google Search AI added app integrations, NVIDIA launched Nemotron 3 Embed, and OpenAI brought PR reviews to Codex. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Microsoft CEO Satya Nadella called Anthropic's Fable "editorially controlled," Kimi K3 became the world's No. 3 AI model, and the EU ordered Google to open Android and Search. And more. Read on for Byte Sized News.

💪 Leverage AI: Kimi K3 just reshaped the AI model race, while Claude Desktop and ChatGPT Work both got major upgrades. Here are the seven AI features worth your attention this week. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Thinking Machines released its first AI model, OpenAI unveiled the Codex Micro, and OpenAI introduced GPT-Red. And more. Check it here!

Ep 821: Claude Desktop Gets Upgrade, New Open Source Model Shocks, ChatGPT Desktop Gets Better and 7 More AI Features You Can Use Today

Is Kimi K3 the shocker of 2026?

Could be.

Now, we have a new (soon to be) Open Model that’s competing with Fable 5 and GPT-5.6, a feat few would have believed possible.

And that was the only new and important drop this week in AI.

Claude brought useful browser to the desktop, ChatGPT made a big fix to how ChatGPT Work works and Google rolled out avatars that could change content creation.

Don’t miss our Friday Features show, where we recap the most important AI updates and features you can use today.

Also on the pod today:

• Claude desktop adds built-in browser 🖥️
• ChatGPT desktop gets full sync 🔄
• OpenAI’s universal search upgrade 🔍

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – River is An AI that demos and closes your deals, Graft AI gives AI agents a reliable way to use legacy software by turning real interfaces into stable tools, Cito is a Semantic search over academic papers.

Google Search AI — Now you can link your favorite apps like Instacart, Canva, and YouTube Music directly in Google Search's AI Mode.

Nemotron 3 Embed — NVIDIA just dropped Nemotron 3 Embed, their new open embedding models that set a retrieval benchmark record and offer flexible deployment for enterprise AI.

NVIDIA Cosmos 3 Edge — Nvidia’s new Cosmos 3 Edge model could spark the next big wave in robotics and physical AI.

Microsoft and 3M — Microsoft is teaming up with 3M to boost AI data center speed and reliability using new optical connectors

Google Vids Upgrades — Google Vids just got a boost with Gemini Omni and personal avatars, letting you create and edit videos using simple prompts, plus star in them with your own digital double.

China and AI — China’s Xi Jinping pushed for global teamwork on AI and warned about the risks of losing control.

OpenAI PR Chat — OpenAI's new PR Chat lets you review, edit, and accept pull requests without ever leaving Codex.

1Password and Claude — 1Password just launched a way for Claude to handle your logins without ever seeing your passwords.

Meta AI Safety — Meta wants AI to flag risky teen behavior on Instagram, but privacy experts say it could make things even messier.

Deepmind and Isomorphic Labs — Google DeepMind and Isomorphic Labs are teaming up to use AI for faster outbreak detection, safer drug design and stronger biosecurity.

Grok Automations — Automate your busywork with Grok: just describe the task once, pick when it runs, and let Grok handle it from there.

Roblox AI — Roblox is rolling out Build, a mobile-first AI tool that lets anyone create games from text prompts right in the app.

1. Nadella Faults Anthropic’s Fable Limits as Microsoft Pushes AI Choice 😱

According to CNBC, Microsoft CEO Satya Nadella told employees Wednesday that Anthropic’s Fable model can feel overly controlled when it rejects user requests.

His remarks arrive as Microsoft is pushing businesses toward cheaper, customizable AI models that can run on their own data instead of relying on a small group of major model providers.

2. Kimi K3 Ranks Third on Artificial Analysis, Open Weights Due July 27 🧑‍🎨

Kimi has launched K3, a 2.8-trillion-parameter multimodal model with a one-million-token context window, and says its full weights will be released by July 27.

According to Artificial Analysis, K3 scores 57, placing it third behind Claude Fable 5 and GPT-5.6 Sol, while outperforming several major rivals on agentic and knowledge-work tests.

3. EU Orders Google to Open Android and Search Data to AI Rivals 🔍

EU regulators have set new Digital Markets Act requirements that will force Google to give competing AI assistants access to 11 Android features and share certain anonymized search data with rivals such as OpenAI.

The Android changes are due with the next version in July 2027, while the search-data rules begin in January 2027, aiming to make it easier for alternative AI and search services to compete with Gemini and Google Search.

4. Google’s Gemini 3.5 Pro reportedly misses its June launch target 🎯

According to Bloomberg News, Google’s flagship Gemini 3.5 Pro model is running months late as the company works to strengthen its coding performance. The reported setback lands as OpenAI and Anthropic push competing systems, increasing pressure on Google to show that Gemini can keep pace at the high end of the market.

5. Meta’s Muse Spark 1.1 Arrives on OpenRouter for U.S. Developers ⚡

Meta says Muse Spark 1.1 is now available through OpenRouter, giving U.S.-based developers a quicker way to access the model without setting up a separate vendor relationship.

The move matters because OpenRouter acts as a single testing and billing hub for many AI models, putting Meta’s newest offering in front of a large developer audience immediately.

6. OpenAI Restores ChatGPT as the Main Screen After “Work” Backlash 👷

OpenAI has quickly revised its new desktop app after users complained that its core chat feature had been oddly sidelined by the launch of “ChatGPT Work.”

The updated app once again opens directly to the familiar “Message ChatGPT” interface, while Work, Codex, Projects, and recent chats have been moved into clearer secondary positions.

The priciest assumption in your AI strategy died late Thursday night: the one where frontier level intelligence only lives behind closed labs and $200 monthly subscriptions.

A Chinese AI lab kinda just tore that strategy down.

Moonshot AI dropped Kimi K3, a soon to be open weights monster punching up to trade blows with the two best closed models on the planet.

Meanwhile, Anthropic stuffed a full browser inside Claude Desktop, and OpenAI rebuilt ChatGPT Work after fumbling the first version.

Biiiig week.

Skip this one and you're prolly walking into Monday's AI strategy conversation armed with last week's intel.

Today's Everyday AI Friday Features show breaks down all seven updates, each one usable today, so you can be the smartest AI voice in the room by lunch. Let's get into it.

1. Claude Code Finally Browses the Web 🌐

Anthropic added a built-in browser to Claude Code desktop, letting it click, type, and act across external websites while handling a task. Paid macOS and Windows users get a clean browser profile, with approval required for every new site.

Finally.

Developers can keep documentation, bug trackers, dashboards, and execution inside one loop, while security teams retain explicit control over Claude’s external actions.

Try This

Choose one development task that requires documentation, a bug tracker, and a dashboard. Run the entire workflow inside Claude Code desktop and track the tab switching, copy-pasting, and manual handoffs eliminated.

2. ChatGPT Makes Your History Searchable 🔎

OpenAI added universal search across ChatGPT chats, Projects, images, and documents from one sidebar box. It’s rolling out to free and paid users across web, iOS, and Android.

Small? Only if your team enjoys rebuilding work it already created.

Old analyses, contracts, images, and unfinished thinking can now become reusable institutional knowledge instead of digital landfill, cutting the scrolling and expensive rework that punish heavy ChatGPT users.

Try This

Search for one valuable analysis, document, or image your team created months ago. Use it in a current project and compare the retrieval time against how long rebuilding that asset would have taken.

3. Superhuman Prewrites Emails Before You Open ✉️

Superhuman upgraded Auto Drafts to prepare responses and unanswered-email follow-ups before you open the inbox. Business and Enterprise users can set it up on desktop, with drafts syncing across mobile, Gmail, and Outlook.

Anthropic and OpenAI models replace GPT-3.5 underneath, which should make those drafts sound more like the actual user.

This won’t reinvent email, but shifting from writing every response to reviewing ready-made drafts can turn inbox management into a much faster approval workflow.

Try This

Use Auto Drafts for one workday and classify every suggestion as send, light edit, or complete rewrite. Keep it in the workflow only if the first two categories dominate and your average response time drops.

4. Spotify Turns Listening Into a Conversation 🎧

Spotify launched Talk to Spotify, a beta that lets Premium users control and discover content through an ongoing voice or text conversation. It’s rolling out on iOS and Android in the U.S., Ireland, and Sweden.

The advantage is your data.

Spotify already knows your playlists, repeat listens, and listening history, showing how apps with proprietary customer context can create experiences a general chatbot simply can’t copy.

Try This

Ask it to build a work playlist using your actual listening history, then refine the energy, genre, and recency through follow-up requests. Compare the result against manual playlist hunting based on setup time and how often you skip tracks.

5. Google Turns Your Likeness Into Video 🎬

Google added personal avatars powered by Gemini Omni, letting users record their face and voice once, then reuse that likeness in generated videos. It’s available to paid U.S. users age 18 or older and also works inside Google Vids.

The win isn’t replacing every real video.

It’s generating approved B-roll, training visuals, social clips, and executive footage when cameras or schedules become the bottleneck, turning one verified likeness into a reusable creative asset.

Try This

Choose one approved executive, trainer, or spokesperson and generate three short B-roll clips around an existing script. Compare production time, reshoots, and approval revisions against your normal video process.

6. ChatGPT Work Finally Feels Like ChatGPT 🔄

OpenAI updated ChatGPT Work desktop so chats and Projects appear in the sidebar, while Chat and Work history sync across web, mobile, and desktop. Users can switch modes more naturally, while local tasks remain on the computer.

The original setup felt disconnected.

Now teams can move between normal conversations and autonomous desktop work without rebuilding context or hunting through separate interfaces, removing the friction that quietly kills adoption.

Try This

Move one active Project from the web into the desktop app and alternate between Chat and Work while completing it. Count how often you repeat instructions, rebuild context, or hunt for a missing conversation.

7. Kimi K3 Blows Open the Model Race 🧠

Moonshot AI released Kimi K3, a 2.8 trillion-parameter model with native vision and a one-million-token context window. It’s live in Kimi, Kimi Work, and the API, with open weights scheduled for July 27.

Kimi K3 landed third on the Artificial Analysis Intelligence Index and beat closed flagships on several important benchmarks covered in today’s show.

Yeah, this is big.

Enterprise teams can evaluate frontier-adjacent autonomous work without accepting frontier pricing or vendor dependence, although Kimi K3 is far too large for normal consumer hardware.

Try This

Run the same long, multi-file business task through Kimi K3 and the closed model your team currently prefers. Compare accuracy, completion time, required interventions, and cost before brand familiarity picks the winner.

Ep 820: The Most Important AI Model You’ll Probably Never Use That Just Dropped

Everyday AI — Thu, 16 Jul 2026 17:07:12 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Thinking Machines just launched Inkling. Here's why the new American open AI model matters for businesses.Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Suno hack shows alleged scraping, states are shaping new AI safety laws, ChatGPT upgraded search, Windows Search is getting AI upgrades and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Thinking Machines released its first AI model, OpenAI unveiled the Codex Micro, and OpenAI introduced GPT-Red. And more. Read on for Byte Sized News.

💪 Leverage AI: Thinking Machines took the lead on American open source with its first release. But it means more than that. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: The Trump administration launched a new AI cybersecurity hub, OpenAI is reportedly building a smart-home speaker, and New York paused new AI data centers. And more. Check it here!

Ep 820: The Most Important AI Model You’ll Probably Never Use That Just Dropped

You've probably never heard of Inkling.

It's the newest (and first) model from Thinking Machines Labs, and it could very well be a small snowball that picks up major momentum in today's enterprise AI landscape.

If you haven’t heard of Thinking Machines, they’re led by Mira Murati, the former CTO at OpenAI.

The big bet with Inkling?

The future of AI could be using smaller models fine-tuned and optimized for smaller tasks.

Will it work?

Tune in live as we dive in.

Also on the pod today:

• Inkling: new American open model 🇺🇸
• Open source vs proprietary clash ⚔️
• Multimodal model: text, image, audio 🖼️

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – V2Fun is AI-Powered 3D creation, Agently builds a company brain across your whole stack, then spins up agents to handle the work, Crustdata builds your sourcing twin in Claude

US AI Safety — States like California, New York, and Illinois are shaping a national AI safety standard, with federal and global frameworks on the horizon.

ChatGPT Search Upgrade — ChatGPT's search just got a big upgrade, letting you find chats, projects, and files way faster from one spot.

Anthropic Misalignment — AI agents are already showing signs of going rogue in simulated high-stakes situations, from secretly sabotaging code to coaching humans on whistleblowing.

Codex in Chrome — Codex in Chrome automates your workflow by pulling context from everywhere and flagging what needs attention.

AI Hacker — A hacker leaked Suno’s secret training data, revealing the AI music tool scraped millions of songs from YouTube, Deezer, and more.

NVIDIA and Japan — Japan’s top tech, banking, and medical giants are taking AI to the next level with NVIDIA, from smarter robots and healthcare breakthroughs to faster financial data crunching and next-gen gaming.

Anthropic Usage Reset — Anthropic surprised users by resetting rate limits early, sparking questions about whether manual resets might be coming.

AI in Bio — Google DeepMind is rolling out AI to help spot and stop biological threats. Find out how their new bioresilience push could change outbreak response.

Spotify Pulls AI Slop — Spotify just pulled a bunch of "AI slop" tracks and now wants artists to be upfront if they use AI in their music.

Windows Search AI — Windows Search is getting smarter with AI, making it faster and more accurate.

1. Thinking Machines Releases Open-Weight Inkling Model 🫟

Thinking Machines Lab on Wednesday unveiled Inkling, its first in-house AI model and a rare open-weight challenger to the closed systems sold by OpenAI, Anthropic, and Google.

The 975-billion-parameter mixture-of-experts model activates about 41 billion parameters per task and can process text, images, audio, and video, although it currently responds only in text.

2. OpenAI and Work Louder Launch $230 Codex Micro Controller for Agent Workflows ⌨️

OpenAI and Work Louder have opened pre-orders for the Codex Micro, a compact keyboard built to give Codex users physical controls for managing AI coding tasks.

The $230 device uses RGB-lit agent keys, a joystick, command buttons, and a dial for changing reasoning intensity, turning commonly used software actions into tactile shortcuts.

3. OpenAI Unveils GPT‑Red to Hunt Prompt-Injection Flaws 🟥

OpenAI says it has trained GPT‑Red, an internal automated red-teaming system designed to find prompt-injection weaknesses faster than human testing alone can manage.

The announcement matters because AI agents increasingly handle outside emails, websites, files, and tools, creating more chances for hidden instructions to hijack their actions.

4. Perplexity Launches SPACE🚀

Perplexity has introduced SPACE, a new platform for running AI agents that need to edit files, use tools, and retain work over extended tasks without leaving sensitive data exposed.

The system uses short-lived isolated environments, while rolling snapshots let agents pause, resume, or branch work later without keeping one sandbox alive indefinitely.

5. Google Renames NotebookLM to Gemini Notebook, Adds Cloud Computing 📓

Google has renamed NotebookLM to Gemini Notebook, broadening the research tool’s reach across Gemini, Search, and its standalone app. The timely update gives notebooks a secure cloud computer that can write and run code for source-grounded data analysis, initially for select paid Google and Workspace users.

Google says the feature will reach Pro web users in coming weeks, while notebook syncing is already available in the Gemini app and Search integration is next. The move signals Google is turning a once-specialized note tool into a more connected research hub, with the name change serving as the smallest part of the story.

6. Moonshot AI Rolls Out Kimi K3 With 1 Million-Token Context 🌝

Moonshot AI is reportedly rolling out Kimi K3 across its web platform, apps, command-line tool, and API, led by a claimed 1 million-token context window that could let it handle book-length material or very large software projects in one go.

The report says K3 also uses a Mixture-of-Experts design with an estimated 2 to 3 trillion parameters and can coordinate as many as 300 sub-agents for complex tasks.

Using your smartest AI model for every task is becoming the clearest sign your AI strategy is immature.

Thinking Machines Lab’s Inkling isn’t trying to be number one.

Good.

It’s a 975-billion-parameter, multimodal, American open-weights model built to be customized, which makes it strategically useful in a way benchmark trophies can’t capture.

Enterprise AI is moving from one premium model for everything to a portfolio of intelligence: credible open models for leverage, tuned specialists for repeatable judgment, and frontier models only when the work genuinely demands them.

Companies that figure this out first can lower costs, loosen vendor dependence, and hardwire their best internal judgment into models competitors can’t buy off the shelf.

That’s what we tackled today on Everyday AI: why Inkling can reopen procurement even if you never deploy it, how frontier models are making fine-tuning practical again, and why model routing is about to become an executive discipline.

1. Reopen procurement with American open weights 🔓

For years, the open-model conversation has mostly pointed toward Chinese labs, keeping many enterprises with government contracts, geopolitical concerns, or board resistance from seriously considering the category.

Inkling could change the buying conversation. It doesn’t need to beat Claude Fable 5 or GPT-5.6 Sol to give companies another path for customization, enterprise distribution, cost control, and vendor leverage.

Procurement gets leverage even without deployment. Vendors negotiate differently when the alternative is real.

Try This: List the workflows your team has kept on closed models because open models couldn’t clear governance or procurement. Reopen the top three with security, legal, and finance, then compare cost, control, deployment requirements, and vendor lock-in.

Don’t switch for novelty. Build enough optionality that your current provider has to keep earning each workflow instead of inheriting it forever by default.

2. Turn proprietary judgment into infrastructure 🔥

Could fine-tuning be making its comeback?

Tinker manages the training infrastructure, while today’s frontier models can generate data, evaluations, code, and failure analysis that help companies shape smaller specialists without building a research lab first.

Bridgewater AIA Labs used Tinker to tune Qwen3-235B for recurring financial judgment tasks. The custom model beat the best frontier model tested and delivered a 13.8x reduction in inference cost per task.

Cheaper output is nice. Turning the judgment your best people struggle to explain into repeatable infrastructure is where this gets dangerous for competitors.

Once that judgment becomes repeatable, it can move through the business without waiting for the expert, burning premium tokens, or depending on another generic prompt.

Try This: Choose one high-volume decision that your top employees make repeatedly and can score after the fact. Gather real examples, define what good looks like, and benchmark a frontier model before testing whether fine-tuning can improve consistency or cost.

Start narrow. A model trained for one valuable judgment beats a vague company-wide fine-tuning experiment.

3. Route every workflow by business stakes 🚀

Put procurement flexibility and fine-tuning together, and the one-model-for-everything era starts looking kinda silly.

Stable, repeated work can move to economical models, proprietary judgment can move to tuned specialists, and ambiguous or high-risk work can escalate to the frontier.

The executive question changes from which model is smartest to which model is enough for this job. That shift is how AI usage becomes sustainable across the business instead of turning into a very expensive free-for-all.

Score workflows by stakes, volume, privacy, and proprietary judgment. Those four variables tell you where cheaper models are safe, where tuning could build an advantage, and where frontier intelligence still earns its price.

Try This: Map 10 recurring workflows and assign each a default model, an escalation model, and a measurable quality floor. Route the lowest-risk, highest-volume workflow first, then compare cost and output before expanding.

Your competitors don’t need a smarter model to beat you. They just need to stop wasting theirs.

Ep 819: ChatGPT Work: What’s New, Who It’s For and How to Use It

Everyday AI — Wed, 15 Jul 2026 15:02:14 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: OpenAI's ChatGPT Work makes Codex easier to use without removing its core AI agent capabilities. Here's what businesses need to know. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Claude is giving teachers free AI tools, Anthropic is backing tougher AI safety laws, and Elon Musk says X will open source its code. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: The Trump administration launched a new AI cybersecurity hub, OpenAI is reportedly building a smart-home speaker, and New York paused new AI data centers. And more. Read on for Byte Sized News.

💪 Leverage AI: ChatGPT Work makes OpenAI's AI agents easier to use by hiding the complexity behind Codex. Here's how to put them to work across your business. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Grok reportedly sent entire Git repos to xAI servers, Meta is facing an AI-related layoffs lawsuit, and OpenAI's GPT-5.6 arrived on Amazon Bedrock. And more. Check it here!

Ep 819: ChatGPT Work: What’s New, Who It’s For and How to Use It

OpenAI's new ChatGPT Work is on fire. 🔥

The company said they've added 2 million users in the past two days alone.

Wanna know a secret about it though?

(It's kinda not new at all. It's just Codex in Chat clothing.)

Regardless, millions of new users are realizing for the first time what it means to have an agent that can work for them 24/7. And even if you're not a technical builder, it's pretty easy to set up a swarm of agents to help with any kind of work.

We show you how.

Also on the pod today:

• ChatGPT Work vs Codex explained 🤔
• Desktop vs web differences 🌐
• Remote mode from your phone 📱

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – ClawTeams is An IM-native AI staff collaboration platform, Branda is On-brand ads generated in seconds, Vocalvia Turns any document into an editable podcast episode.

Claude for Teachers — Claude for Teachers is giving US K-12 educators free access to AI-powered lesson planning and classroom tools, all mapped to state standards.

Anthropic State by State — Anthropic is pushing for the toughest AI safety laws yet, backing strict audits and risk checks in several states.

George Lucas AI — George Lucas calls out Hollywood for letting focus groups and fans steer movies, saying real storytelling should come from passionate filmmakers.

Superhuman AI Emails — Superhuman’s new AI drafts emails that actually sound like you, not a bot.

Meta Muse Spark 1.1 Arena — Meta’s Muse Spark 1.1 just debuted in Agent Arena, landing at #17 and beating Gemini 3.1 Pro.

Codex Micro — Work Louder’s Codex Micro lets you monitor ChatGPT agents with physical RGB keys.

Sesame Calendar — Sesame is adding Calendar and Email connectors for its iOS app, letting AI handle more for you.

Spotify Controversial AI Feature — Spotify’s new conversational feature lets you shape your music, ask questions, and explore your listening history just by talking or typing

X Open Source — Elon Musk says X will open source its entire codebase after a security review, promising full transparency.

Gemini in Chrome UK — Gemini is now built into Chrome for UK users, letting you chat with an AI assistant to summarize pages or draft emails right in your browser.

Apple AI — Apple is eyeing tech that shrinks massive AI models to fit on your iPhone, promising faster, more private features.

ChatGPT Cycling — After losing his leg as a child, Vishal used ChatGPT to train, plan nutrition, and tweak his prosthetic. Now he's set to represent India at two major para cycling championships.

1. Trump Administration Launches GOLD EAGLE Cyber-Patching Hub 🦅

The Trump administration has announced GOLD EAGLE, a new federal-industry clearinghouse designed to speed up the detection, verification, and patching of cybersecurity flaws affecting critical infrastructure.

Created under a June 2 executive order, the effort brings Treasury, DHS and CISA, the Department of War, open-source partners, and major infrastructure companies into a shared response system. The central bet is that frontier AI can help defenders sort urgent risks faster while cutting repeated scanning work.

2. OpenAI Reportedly Targets Smart-Home Speaker With Humanlike AI Personality 🔊

According to Bloomberg, OpenAI is developing a mobile, screen-free smart speaker that could mark its first major move into consumer hardware ahead of a potential IPO.

The device is designed to manage smart-home tasks, media, messages, and ChatGPT features while learning its owner’s habits and drawing on personal data such as email. Its moving mechanical parts and conversational style are meant to make it feel less like a gadget and more like a household companion.

3. Apple Tests PrismML’s Tiny iPhone AI Models 📱

According to CNBC, Apple is evaluating PrismML’s technology, which the startup says can compress a 27-billion-parameter AI model from about 54 GB to under 4 GB so it can run directly on newer iPhones.

The news lands just as Apple opens an iOS 27 beta featuring its revamped Siri, underscoring its push to make the assistant faster and more private without sending every request to the cloud. PrismML claims its compressed models use far less memory, respond much faster, and consume less power, though they give up some factual accuracy.

4. New York Pauses New 50-MW AI Data Centers for Up to a Year ⚡

Gov. Kathy Hochul has signed an executive order temporarily blocking new large-scale data centers that use at least 50 megawatts of power, citing pressure on the grid, water resources, and household utility bills.

The move makes New York the first state to impose a statewide pause of this kind, giving officials time to set rules for local siting, cleaner power, and consumer protections.

5. Publishers and Authors Sue Google Over Gemini’s Book Training 📚

A new class action lawsuit accuses Google of using copyrighted books from Google Books and Google Play to train Gemini without permission, according to TechCrunch.

The publishers and authors also claim Google altered or removed copyright information to hide the alleged use of protected material. The filing lands as courts weigh whether AI training qualifies as fair use, with early California rulings favoring AI companies but leaving major questions unresolved.

6. IBM Plunges 23% After Z17 Mainframe Miss Raises AI Spending Doubts 💸

IBM shares tumbled more than 23% at Tuesday’s open after its second-quarter profit and revenue fell short of Wall Street expectations, putting the company on track for its worst day in decades.

CEO Arvind Krishna said weaker-than-expected sales tied to the Z17 mainframe and its transaction-processing software were the central problem, despite the system’s AI-focused pitch. IBM also said customers redirected late-June budgets toward servers, storage, and memory ahead of expected price increases, leaving less room for its software and infrastructure purchases.

One word kept nontechnical workers outside OpenAI’s desktop agent: code.

ChatGPT Work changed the label, simplified the view, and two million users showed up in 48 hours.

Pay attention.

Codex already had computer access, browser use, files, apps, automations, long-running goals, and parallel subagents. The breakthrough was making all of that feel like normal work.

The competitive gap is simple: one team babysits a chatbot while another directs several agents across real workflows.

That gap can get ugly fast.

On today’s Everyday AI, we broke down the packaging play, the desktop-versus-web decision, and the management system leaders need before agentic work spreads faster than their controls.

1. OpenAI made Codex feel usable ⚡

ChatGPT Work on desktop is largely Codex behind a friendlier front door. Work mode hides the code-heavy detail; Codex keeps the diffs, changed files, and technical visibility.

Same engine, better invitation.

OpenAI removed the developer-first framing and pointed users toward finished documents, presentations, analysis, dashboards, and other business deliverables.

That should make leaders rethink every AI rollout built around licenses, feature tours, and vague encouragement to experiment. Access means little when employees can’t connect the tool to the work already on their plate.

Nah.

Buying capability and hoping adoption follows ain’t a strategy.

Try This: Rebuild one underused AI rollout around three finished outputs, not a feature list.

Give finance a variance brief, sales a researched account plan, or operations a live tracker. If employees still need technical jargon to get started, simplify the doorway again.

2. Move ChatGPT Work to desktop 🚀

ChatGPT Work on the web gives casual users a familiar place for longer jobs. Desktop creates the bigger shift because it can use the browser, computer, local files, apps, connected tools, Remote, and scheduled tasks.

One instruction can analyze beehiiv performance, inspect months of context, build a mobile-friendly site, publish it, and attempt the final handoff. That collapses a scattered chain of knowledge work into one directed outcome.

Remote lets you start, steer, or check that work from your phone while the machine handles execution elsewhere.

Add queues, loops, long-running goals, and subagents, and one employee can supervise several workstreams instead of waiting for one thread to finish.

This changes the job.

Your people move from prompting an assistant to directing specialists, which only works when the workflow, context, and finish line are clear enough to avoid constant rescue.

Try This: Choose one weekly process that crosses at least three systems and ends in a shareable deliverable.

Map the approved sources, output format, review point, and final action. Once the workflow holds up, split research, production, and QA across subagents.

3. Manage agents before they multiply 🔥

More capable agents create a management problem fast.

They can run for hours or days, rewrite files, use your computer, queue follow-ups, and create their own specialist teams. Goals define the finish line.

Plan mode exposes the approach, steering corrects the work midstream, and AGENTS.md files preserve the rules across the project.

Permissions still matter because approval friction is real and unrestricted computer access is reckless.

The companies that scale this well will decide what agents may read, write, publish, send, and spend before dozens of autonomous tasks create an expensive, untraceable mess.

Yeah, governance sounds less exciting than a viral launch. It also decides whether ChatGPT Work creates dependable capacity or polished mistakes moving at machine speed.

Try This: Write a one-page operating brief for the first workflow: goal, approved folders, connected apps, forbidden actions, output standard, reviewer, and usage ceiling.

Turn those decisions into an AGENTS.md file and require review before external messages, publishing, or irreversible actions. Then let the agent run.

Ep 818: AI Just Went Multiplayer. Slack Is Where It All Comes Together

Everyday AI — Tue, 14 Jul 2026 14:10:10 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Sup y’all 👋

On Wednesdays, we do hands-on demos of some of the latest/greatest AI features you can use.

Quite a few on the plate.

See ya tomorrow

✌️

Jordan

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: AI is becoming a team sport, and Slack wants to be where people and AI agents work together. Here's how that changes the way businesses get work done. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Sam Altman and Elon Musk are clashing over space data centers, Siri is getting smarter on Apple Watch, and AI protests have reached San Francisco. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Grok reportedly sent entire Git repos to xAI servers, Meta is facing an AI-related layoffs lawsuit, and OpenAI's GPT-5.6 arrived on Amazon Bedrock. And more. Read on for Byte Sized News.

💪 Leverage AI: Slack is becoming the place where AI agents, company knowledge, and teams work together. Here's how businesses can use it to get more from AI. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Report: Apple sued OpenAI over alleged hardware trade-secret theft, OpenAI's No. 2 executive is stepping down, Big Tech has added $350 billion in debt to fund AI data centers and more. Check it here!

Ep 818: AI Just Went Multiplayer. Slack Is Where It All Comes Together

Of course you need data to fuel your AI. You know what's just as helpful though? 🤔

Decisions.

And the conversation and rationale behind it.

As agents become more mainstream, the required structured data only goes so far.

↳ What about how your team makes decisions?

↳ Or the real reasons why one RFP landed and one flopped?

↳ Or where Deborah in accounting keeps saving that KPI sheet and if it's actually approved or not?

(Cmon Deborah.)

The big advantage moving forward is giving your agents access to the rationale and context of your team's conversations.

And that's the big unlock that Slack just unveiled.

Also on the pod today:

• Inbox is “death” for AI productivity ✉️
• Slackbot: fastest-growing Salesforce feature 🚀
• AI assistants reading team conversations 🤖

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

1. Grok Build Reportedly Sent Entire Git Repos to xAI Servers ✌️

A July 12 investigation by researcher Cereblab alleges that xAI’s Grok Build uploaded full Git repositories, including untouched files and commit history, to a Google Cloud bucket during routine code-analysis requests.

In one test, more than 5.1 GB reportedly left a repository even though the task required only about 192 KiB, raising fresh concerns for teams with credentials or private code in local projects.

2. Meta Sued Over Claims AI Helped Target Workers on Protected Leave for Layoffs 😱

Twenty-six current and former Meta employees filed suit Monday, alleging the company used AI-driven monitoring and performance tools to help select workers for a recent 10% workforce reduction.

The complaint says employees who took or sought protected medical, family, disability, or pregnancy-related leave were unfairly disadvantaged because the system weighed productivity data without properly accounting for time away.

3. OpenAI’s GPT-5.6 Sol, Terra, and Luna Arrive on Amazon Bedrock 🧑‍💻

Amazon Bedrock has added OpenAI’s GPT-5.6 model family, giving AWS customers three options aimed at deep reasoning, everyday production work, and fast high-volume tasks.

The announcement matters because it places OpenAI’s newest claimed flagship capabilities inside AWS infrastructure, where companies can keep workloads in selected regions and apply existing AWS security controls. Bedrock also adds prompt caching for repeated agent context, a feature designed to lower costs when applications make many related model calls.

4. Perplexity Adds Self-Improving Memory, Faster Models, and One-Click Website Publishing 🖱️

Perplexity has rolled out a major Computer update that gives its AI agent a persistent, source-linked memory designed to carry lessons, files, and prior decisions across tasks.

The launch also adds faster access to Opus 4.8, Claude Fable 5 for coordinating complex work, and the ability to switch orchestrator models midway through a task. Users can now publish Computer-built websites to Perplexity-hosted links or Vercel, while organizations gain tighter sharing controls and per-model usage analytics.

5. Microsoft’s Nadella Warns Enterprises About AI Data Exposure 🛡️

In a newly published blog post, Nadella argues that companies using proprietary AI models may be paying twice: once in fees and again by exposing valuable internal knowledge through prompts, corrections, and agent activity.

He says that information can teach model providers how a business works, raising concerns that vendors could gain insight into their customers’ competitive advantages.

6. DeepMind’s Hassabis Urges US-Led AI Security Testing Body 👮

Google DeepMind chief Demis Hassabis is calling for urgent US leadership on a new standards body to test the most powerful AI models for national-security threats, following Washington’s recent export ban on Anthropic’s top systems.

According to the Financial Times, he wants a Finra-style watchdog to assess models worldwide, including whether their safety controls can be bypassed.

Private AI chats are making employees smarter and companies dumber.

Every prompt, correction, and decision trapped with one person creates a tiny productivity win that never compounds. Meanwhile, AI-native competitors are turning shared conversations into reusable intelligence.

That’s the multiplayer AI problem we unpacked on today’s Everyday AI with Ryan Gavin, Chief Marketing Officer at Slack.

The real opportunity isn’t another agent. It’s connecting your people, company context, and entire software stack so work moves faster without everyone becoming an AI traffic cop.

1. Stop scaling private AI wins 🔥

Most companies still deploy AI like a solo tool. One person gets faster, but the team never sees the prompt, learns from the correction, or improves the process.

Ryan said the collaboration layer can represent two percent of AI spend while unlocking the value of the other 98 percent.

That’s the gap. AI output isn’t enterprise productivity until a team can challenge it, improve it, assign it, and ship it.

Try This

Pick one cross-functional workflow and run it in a shared Slack channel for 30 days. Keep the prompts, agent outputs, corrections, decisions, owners, and results together.

If the learning still dies with one person, your AI strategy is still single-player.

2. Capture how decisions happen ⚡

Slack’s real AI asset isn’t messaging. It’s years of decision logic, customer context, failed attempts, and team judgment hiding inside conversations.

Ryan said he uses Slackbot to scan hundreds of channels, find blockers, track progress against goals, and show where his attention matters most.

That changes management fast. Leaders can stop collecting status and start fixing the work.

The bigger implication? Communication policy is now AI architecture. Every useful decision buried in a DM weakens the next answer your company gets.

Try This

Create one daily Slackbot briefing for blockers, goal progress, and decisions requiring executive judgment. Then move repeatable decisions into shared channels and close major threads with the owner, rationale, and next step.

Better context today becomes better management leverage tomorrow.

3. Make the software disappear 🚀

Agent sprawl is about to become app sprawl with better branding. Employees may have dozens of agents available and still have no clue which one to use.

Ryan said Slackbot can call multiple agents and read or write across permissioned systems through one conversation.

That’s the headless opportunity. Employees shouldn’t need to understand every tool, tab, or back-end system to get work done.

The competitive advantage won’t come from owning more software. It’ll come from hiding the complexity better.

Try This

List the 10 to 12 systems a role touches every day, then identify the five most common actions across them. Route the top three through one permissioned conversational layer and keep human approval for high-consequence changes.

Measure completed work, handoff delays, and time spent inside apps. Not chatbot activity.

Ep 817: ChatGPT’s 5.6 Sol, Grok and Meta bounce back and OpenAI’s biggest week ever? And more AI News That Matters

Everyday AI — Mon, 13 Jul 2026 15:39:48 +0000

👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: OpenAI launched GPT-5.6 and ChatGPT Work, while Meta rolled out new AI models and its first AI chip. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: OpenAI announced ChatGPT is back on WhatsApp, Anthropic extends Fable access in subscriptions for a week, and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Report: Apple sued OpenAI over alleged hardware trade-secret theft, OpenAI's No. 2 executive is stepping down, Big Tech has added $350 billion in debt to fund AI data centers and more. Read on for Byte Sized News.

💪 Leverage AI: OpenAI reclaimed the AI lead with GPT-5.6, while Microsoft, Meta, and SpaceXAI showed the next battle is cost—not just capability. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: OpenAI is shutting down Atlas, GPT-5.6 is coming to Microsoft 365 Copilot, Microsoft is bringing more AI to Windows security and more. Check it here!

Ep 817: ChatGPT’s 5.6 Sol, Grok and Meta bounce back and OpenAI’s biggest week ever? And more AI News That Matters

Anthropic had a really bad week. 🥺

First of all, Claude Fable 5 got knocked off at the top by OpenAI's GPT-5.6 Sol.

Then?

Both Meta and Grok came for Anthorpic's API coding pie.

Yikes.

But in between the fights for tokens, we also go some big name lawsuits, weird study results and AI that talks to you more like a person than a robot.

Also on the pod today:

• Anthropic’s brutal benchmark losses 📉
• OpenAI launches GPT Live voice 🗣️
• Apple sues OpenAI over secrets ⚖️

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – FetchSandbox is The memory graph that lets your AI ship integrations without burning real APIs, Flowing is AI writing grounded in your research library, Layoutr is The simple floor plan tool anyone can use

Mac Chip Upgrades — Apple’s new M6, M7 and M8 chips are all about powering AI across Macs.

Meta Glasses for Blind Veterans — Meta and the Blinded Veterans Association are giving Ray-Ban AI glasses to blind veterans, helping them regain independence and connect more easily.

Mini Data Centers — Sunrun wants to pay solar customers to host AI computing units in their homes, turning neighborhoods into mini data centers.

Claude in-app Browser — Claude Code desktop now has an in-app browser, letting you pull up and interact with any site or doc right inside the app.

ChatGPT on WhatsApp — OpenAI announced that its popular message ChatGPT on WhatsApp is back.

Seadream 5.0 Pro Arena — Seedream 5.0 Pro shot up from #11 to #2 on Arena’s Multi-Image Edit leaderboard, marking a major upgrade.

Claude Fable 5 Extended — Claude Fable 5 access is extended on all paid plans, with higher weekly rate limits until July 19.

ChatGPT 5-hour Limit Removed — ChatGPT just Removed its 5-hour limit for Plus, Business, and Pro users. Find out what this means for your workflow and why efficiency upgrades are on the way.

GPT-5.6-Sol Arena — OpenAI’s GPT-5.6-Sol just tied Claude Fable 5 for the top spot in Code Arena: Frontend, matching its performance but at nearly half the cost.

1. Apple Sues OpenAI Over Alleged Hardware Trade-Secret Theft 👾

Apple filed suit Friday against OpenAI, its acquired hardware startup io Products, and former Apple executives Tang Tan and Chang Liu, alleging coordinated theft of confidential information tied to future ChatGPT hardware.

The complaint claims Tan used Apple supplier, manufacturing, and design knowledge to recruit staff, while Liu allegedly accessed Apple systems after leaving and downloaded more than 1,000 pages of internal material. Apple says OpenAI’s growing hardware effort, which works with several Apple suppliers and employs hundreds of former Apple personnel, may be benefiting from information it did not earn independently

2. OpenAI No. 2 Fidji Simo Steps Down Amid Executive Departures 👋

Fidji Simo, OpenAI’s second-in-command, is leaving her full-time role after an extended medical leave, adding pressure to a company already dealing with a string of senior exits.

According to The Wall Street Journal, Simo said her worsening chronic condition requires a longer recovery, though she will remain a part-time adviser.

3. Meta Pulls Instagram AI Image Tool After Three-Day Privacy Backlash 🖼️

Meta removed its new Instagram-based image generator Friday, just three days after launch, following sharp criticism that it let users create images of people from public accounts without clear prior consent.

According to Reuters, public Instagram profiles were included by default, forcing users to find and activate an opt-out setting. SAG-AFTRA called the approach unacceptable and welcomed the shutdown, underscoring growing concern over how social-media photos can be reused by generative tools.

4. Big Tech Adds $350 Billion in Debt for AI Data Centers 🏗️

According to Bloomberg, Alphabet, Amazon, Meta, Microsoft, and Oracle have collectively added about $350 billion in debt over five years to fund the AI data-center buildout.

Amazon’s reportedly cool reception for a $25 billion bond sale this week suggests investors are starting to test how much borrowing they will support, even from the industry’s richest companies.

5. Anthropic Leases 16-Story Manhattan HQ as Airbnb Buys Gramercy Building 🏙️

Anthropic is taking over all 16 floors at 330 Hudson Street, a major New York expansion that could house 1,700 desks as the AI company aims to surpass 1,000 local employees by year-end.

At the same time, Airbnb has bought a six-story Park Avenue South building for $81.5 million, giving its 600-plus New York-area staff a permanent hub despite its long-running fight with the city over short-term rental rules.

6. OpenAI’s GPT-5.6 Wins Big Battle in Frontend Design 🏆

OpenAI’s GPT-5.6 Sol has taken first place on Design Arena with a 1353 Elo rating, moving ahead of Anthropic’s Claude Fable 5 in frontend design, an area where OpenAI had openly worked to improve.

The result is a notable shift: OpenAI’s earlier models were often criticized for generic, poorly polished interfaces, and the company’s own frontend guidance warned that models could fall back on repetitive design habits without detailed direction. Design Arena’s blind voting focuses on how finished web interfaces look and feel, making this less about passing code tests and more about whether users actually prefer the result.

Anthropic didn’t just lose the benchmark crown this week.

It got squeezed from three directions.

OpenAI came in with GPT-5.6 at roughly half the price of Claude Fable 5. SpaceXAI and Meta introduced cheaper coding alternatives.

Microsoft started routing everyday work through its own models.

Meanwhile, Apple sued OpenAI, and voice AI finally stopped feeling like a walkie-talkie.

Welp.

The AI leaderboard, cost curve, and product experience all shifted at once.

Here are the seven stories that matter, in the order they appeared.

1. Apple sues OpenAI while their AI partnership stays intact ⚖️

Apple and OpenAI just entered the most awkward phase of their relationship.

According to reports, Apple sued OpenAI on Friday, alleging its consumer hardware team obtained trade secrets through former Apple employees Tang Yew Tan and Chang Liu. Apple says OpenAI encouraged recruits to share confidential information and avoid scrutiny during hiring.

OpenAI denies wrongdoing and says it is building its own technology.

That would be messy enough on its own.

But Apple still integrates ChatGPT, while OpenAI is developing hardware that could eventually compete with Apple’s core interface. Elon Musk then used the filing to renew his public attacks on Sam Altman, though neither man is a party to the case.

The lawsuit does not prove Apple’s allegations. It does put OpenAI’s hardware plans and one of Apple’s most important AI partnerships under a very bright spotlight.

What it means: This is less about today’s ChatGPT integration and more about who controls the next consumer interface.

For business leaders, the key signal is partnership fragility. Two companies can remain commercially connected while preparing to compete, litigate, and replace one another at the same time.

2. GPT-Live-1 makes ChatGPT voice feel less like a walkie-talkie 🎙️

Voice AI finally stopped acting like a walkie-talkie.

OpenAI launched GPT-Live-1 as ChatGPT’s default voice model for paid plans, with limited free access as well. Unlike earlier voice modes, it can listen and speak at the same time, handle interruptions, stay quiet while someone thinks, and decide when a tool is needed.

The bigger shift is delegation.

Harder questions can move to another GPT model in the background without ending the live conversation. That keeps the interface responsive while deeper work continues.

Video and screen sharing are expected to follow.

The practical example was stronger than the demo reel: opening a 60-page document, discussing it during a drive, asking the model to keep probing based on each answer, then using the transcript afterward.

That is a much more useful version of voice.

What it means: Voice is becoming a working interface, not just a hands-free chatbot.

The useful workflow is document-based thinking: load the source, talk through it, let the model ask adaptive follow-ups, and reuse the transcript. That turns commute time or walking time into structured analysis instead of casual conversation.

3. Grok 4.5 attacks Anthropic’s coding business on price 💸

Anthropic’s coding moat just got a lot more expensive to defend.

SpaceXAI and Cursor jointly trained and released Grok 4.5, giving Cursor a model shaped around the coding platform’s own workflow. That matters because Cursor controls the application and routing.

Users do not need to choose Grok as their general chatbot for it to win coding volume.

The pricing pressure is blunt.

Grok 4.5 costs $2 per million input tokens and $6 per million output tokens. Claude Fable 5 costs $10 and $50.

SpaceXAI also reported more than four times fewer output tokens than Opus 4.8 on one test.

Grok did not lead every benchmark, but it tied or beat Anthropic models on several coding evaluations. For teams refactoring large codebases and burning enormous token volumes, cheaper can become the deciding feature.

What it means: Enterprise adoption may arrive through a narrow door.

Most employees may never use Grok directly, but software teams could use it heavily behind Cursor or another routed interface. Once workloads consume billions of tokens, model selection stops being a brand preference and becomes a unit-economics decision.

4. Microsoft routes Excel and Outlook prompts through its own MAI models 🔀

Microsoft may be changing Copilot without most users ever noticing.

A Bloomberg report says Microsoft has started using its own MAI models for some Excel and Outlook features. Tens of thousands of weekly prompts are reportedly already running through those internal models, making this a production shift rather than a future experiment.

The likely setup is tiered.

Routine work can go to cheaper MAI models. Harder tasks can still be routed to GPT-5.6, Claude Sonnet 5, or another frontier model.

Microsoft gets lower costs while users keep the same interface.

Quality is the unresolved piece. A recent large government trial found that 44% of participants preferred other generative AI tools over Copilot.

Still, this is bigger than Microsoft. As models become more capable and agentic, they also use more tokens and tools.

Companies will increasingly route work based on difficulty, cost, and required quality.

What it means: The model under your enterprise assistant may change long before the product name does.

Teams should evaluate output quality at the task level, not assume every Copilot response uses the same intelligence. The winning architecture may be a cheap default model with frontier escalation only when the work truly needs it.

5. 69% of surveyed Americans back public ownership of major AI firms 🏛️

The public ownership debate around AI is no longer staying on the fringe.

A Verasight survey of roughly 1,700 U.S. adults found 69% initially supported requiring major AI companies to transfer half their stock into a publicly owned fund.

Support crossed party lines: 82% of Democrats, 77% of independents, and 59% of Republicans backed the idea.

Then Bernie Sanders’ name entered the question.

Overall support fell to 64%, though nearly half of Republicans still supported it. Sanders had introduced related legislation about six weeks earlier.

Let’s be honest. A forced 50% equity transfer faces enormous political, legal, and implementation barriers, and it is highly unlikely to happen.

The more plausible signal is smaller public stakes. As AI companies approach public markets, job disruption grows, and data centers move into local communities, some form of shared financial upside could keep returning to the policy conversation.

What it means: Do not focus only on whether the 50% proposal can pass.

The durable issue is who captures AI’s financial upside. Business leaders should expect public-benefit demands, local resistance, and ownership proposals to become part of the operating environment as AI infrastructure and employment effects become more visible.

6. Meta releases Muse models and prepares its Iris AI chip ⚙️

Meta compressed a year of AI catch-up into three days.

Meta rolled out Muse Image, previewed Muse Video, released Muse Spark 1.1, and opened its first public model API. Muse Image can search, write code, refine its own work, and coordinate with Muse Spark.

Meta also placed it inside Meta AI, Instagram Stories, and selected WhatsApp markets.

One feature immediately backfired.

Meta let users generate images from public Instagram accounts, then removed the capability after backlash. Muse Video remains an early preview with native audio, plus acknowledged gaps in synchronization and fast motion.

Muse Spark 1.1 adds coding, computer use, multi-agent orchestration, and a one-million-token context window. Its preview pricing is $1.25 per million input tokens and $4.25 per million output tokens.

Then came Iris, Meta’s first proprietary AI chip. Production is planned for September, supplementing rather than replacing NVIDIA and AMD GPUs.

What it means: Meta is attacking the stack from both ends.

Muse Spark pressures model pricing while Iris targets compute dependence. For companies that burn tokens, price per performance is becoming more important than a single benchmark win.

Sheesh.

Meta suddenly belongs in serious coding-model and infrastructure conversations.

7. GPT-5.6 and ChatGPT Work put OpenAI back on top 🥇

OpenAI had the best week because it paired model leadership with a usable work product.

This past week, OpenAI introduced GPT-5.6 in three tiers: Sol as the flagship, Terra in the middle, and Luna as the cheaper, faster option. Across major aggregate benchmarks, Sol tied or led and reached the top of Design Arena, directly challenging Anthropic’s perceived lead in front-end design.

Pricing creates the sharper pressure.

GPT-5.6 Sol is listed at $5 per million input tokens and $30 per million output tokens, compared with Claude Fable 5 at $10 and $50.

OpenAI also launched ChatGPT Work, a simpler desktop interface built around the same model family. It can produce finished documents, spreadsheets, presentations, analysis, and websites.

The easiest description is Codex Lite, with a toggle for users who still want pull requests, diffs, and deeper coding workflows.

What it means: OpenAI is competing on capability, cost, and interface at the same time.

Anthropic now faces pressure at the frontier from Sol and at the lower-cost coding layer from Grok 4.5 and Muse Spark 1.1. For everyday teams, ChatGPT Work may matter most because it packages agentic execution into a less technical desktop experience.

Everyday AI

Ep 836: Updated GPT-5.6, A new Cheap Meta Model, Qwen 3.8 released and impressive and 7 more AI updates you can use today

Outsmart The Future

Today in Everyday AI8 minute read

Ep 836: Updated GPT-5.6, A new Cheap Meta Model, Qwen 3.8 released and impressive and 7 more AI updates you can use today

Also on the pod today:

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts | YouTube Podcasts| Amazon Music |

1. Adobe Stuffed 70+ Tools Into ChatGPT 🎨

2. Meta Priced Frontier Coding At 95% Off 💸

3. Alibaba Crashed The Top Five Model Club 🏆

4. Gemini Notebook Finally Went Full Agent 📓

5. Agent Plugins Became The Universal Standard 🔌

6. Cloudflare Gave Away Its Company Brain 🏢

7. MiniMax Open Sourced The Best AI Video 🎬

8. Free ChatGPT Became AI's Best Deal 🎁

Ep 835: Inside Everyday AI: My 9 most Used AI Tools and Workflows

Outsmart The Future

Today in Everyday AI8 minute read

Ep 835: Inside Everyday AI: My 9 most Used AI Tools and Workflows

Also on the pod today:

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts | YouTube Podcasts| Amazon Music |

1. Overnight Codex agents end morning busywork 🔥

2. AI skills turn repetition into owned IP ⚡

3. ChatGPT Voice runs your whole computer 🚀

Ep 834: Gemini Notebook: 7 New Updates and What They Unlock

Outsmart The Future

Today in Everyday AI8 minute read

Ep 834: Gemini Notebook: 7 New Updates and What They Unlock

Also on the pod today:

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts | YouTube Podcasts| Amazon Music |

1. NotebookLM dies, Gemini Notebook goes full agent 🚀

2. One prompt now builds five executive deliverables ⚡

3. The 100% grounding promise just got an asterisk 🔥

Ep 833: RSI Explained: When AI Starts Improving Itself and What It Means

Outsmart The Future

Today in Everyday AI8 minute read

Ep 833: RSI Explained: When AI Starts Improving Itself and What It Means

Also on the pod today:

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts | YouTube Podcasts| Amazon Music |

1. OpenAI's model made itself 80% cheaper 🔥

2. Eight weeks that rewrote the AI timeline ⚡

3. The people building RSI want brakes 🚀

Ep 832: OpenAI’s new Astra model, more AI agents escape sandboxes, AI leaders call for AI pacing and more.

Outsmart The Future

Today in Everyday AI8 minute read

Ep 832: OpenAI’s new Astra model, more AI agents escape sandboxes, AI leaders call for AI pacing and more.

Also on the pod today:

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts | YouTube Podcasts| Amazon Music |

1. OpenAI finds more agents that walked out of the sandbox 🚨

2. Anthropic says three Claude models breached real companies 🕳️

3. OpenAI cuts GPT-5.6 Luna by 80% after Sol optimized itself 💸

4. More than 1,000 AI staffers ask Washington for a brake pedal ✋

5. Amazon guts most of the Nova lineup for one frontier model 🪓

6. OpenAI teases Astra, its next model class, inside a math post 🌌

Ep 831: Chrome adds Some Gemini Spark, Replit Design makes impact, Buzz brings AI Agent Teamwork and 7 more AI Features you Should use Today

Outsmart The Future

Today in Everyday AI8 minute read

Ep 831: Chrome adds Some Gemini Spark, Replit Design makes impact, Buzz brings AI Agent Teamwork and 7 more AI Features you Should use Today

Also on the pod today:

Subscribe and listen on your favorite podcast platform

Listen on:

Spotify | Apple Podcasts | YouTube Podcasts| Amazon Music |

1. Replit Design Launches Free for Every User 🎨

2. ChatGPT's Chrome Extension Reads YouTube and Tabs 📺

3. Meta AI Now Runs Recurring Tasks 🔁

4. Gemini Now Generates Images Inside Google Docs 📄

5. Gemini Spark Now Runs Inside Chrome 🌐

6. Lyria 3.5 Lands in Google Flow Music 🎵

7. Block Launches Buzz for Humans and Agents 🐝

Ep 830: Faster AI Agents, Fewer Human Coworkers: The Overly Productive Future of Managing Agents?

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read

Spotify | Apple Podcasts |
YouTube Podcasts| Amazon Music |

Today in Everyday AI
8 minute read