Good morning.

OpenAI shipped GPT-5.6.
Grok 4.5 also dropped.

One of them had to check with Washington first.

Let’s dive in 👇

🍭 What’s Cookin’:

• OpenAI got the government's blessing (GPT-5.6 is live)

• SpaceXAI dropped Grok 4.5 same day. Born blessed

• Anthropic brought Cowork to your phone

• EU regulators gave 110 banks until Halloween to fix their AI cyber gaps

OpenAI's
🙏 GPT-5.6 Is Out. Ask Washington.

The Bite:

GPT-5.6 is a model family split in three: Sol, Terra, and Luna. It was launched publicly today across ChatGPT, Codex, and the OpenAI API.

The models were first previewed on June 26 to roughly 20 government-vetted organizations, at the Trump administration's request.

The Department of Commerce's Center for AI Standards and Innovation ran additional testing before clearing the broader release.

OpenAI sent technical experts to Washington to answer questions in real time.

Sol is the flagship tier, pricing at $5/$30 per million tokens;
Terra at $2.50/$15;
Luna at $1/$6.

OpenAI said it doesn't believe government access "should become the long-term default".

Snacks:

• Sol scored 96.7% on OpenAI's internal cyberattack evaluation, with ExploitBench performance comparable to Anthropic's classified model Mythos Preview at roughly one third of the cost.

• The 12 days of gating period was nominally voluntary under Trump's June executive order, which bars mandatory preclearance.

• Grok 4.5 launched the same day with no government review period disclosed.

• The August 1 deadline requires the National Security Agency to finalize a formal framework designating "covered frontier models". Whatever that classification produces will define who gets gated next.

Why it Bites:

OpenAI called this voluntary. The executive order bars mandatory preclearance.

No licenses required, no approvals needed, just a polite ask from Washington, and twelve days of quiet.

Voluntary frameworks with consequences aren't voluntary.

When the alternative is a Commerce Department directive like the one that pulled Fable 5 offline for 19 days, "cooperating" becomes the only way.

The executive order was sold as innovation-friendly, with no gatekeepers, no preclearance theater.

But what it produced, in practice, is an informal review process with no published criteria, no defined timeline, and no public record of what was tested or what passed.

OpenAI got 12 days. Anthropic got a shutdown. SpaceXAI got a blink.

The difference between those still hasn’t been explained.

If the framework's goal is safety, the question it still hasn't answered is why it applies to some frontier models and not others.

And if the answer turns out to be relationships, that's not a safety framework.

Pick your own word for it.

Steal This Prompt
🚞 Vintage Posters From a Design Museum

Turn any photo into a bold, vintage two-color illustration that looks ripped straight out of a retro screen-printed poster.

Use it to:

• Transform ordinary photos into gallery-worthy vintage prints

• Give your car, motorcycle, or travel photos retro-cool vibes

• Create bold branding, merch, wallpapers, and social graphics

Workflow:

1. Hit this link: Vintage Dual-Color Art

2. Paste into your AI model

3. Replace the #s with your image details or color preferences

4. Watch it cook into a vintage masterpiece

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🛠️ Timbal AI

One platform for the full AI app lifecycle: agents, workflows, retrieval, observability, and evals without stitching five tools together.

📈 Auriko

Routes each request across providers by token price, latency, and cache behavior, averaging 30% savings.

💬 Toyo

An exec assistant that lives in iMessage and calls your phone, triages your inbox, preps you for meetings, and pulls context from your company tools over text or voice.

🔐 Opper AI

One API key to 300+ EU-hosted models, zero prompt storage by default, and a base URL change away from your existing OpenAI or Anthropic SDK.

🔌 Blocks.ai

PubNub's control plane for connecting AI agents across any framework or provider without opening ports, changing DNS, or touching your firewall.

_{Everything Else}
🧠 You Need to Know

🦾 SpaceXAI Launches Grok 4.5 For Developers And Enterprise
→ The first SpaceXAI model since going public claims Opus-class performance at $2/$6 per million tokens (roughly one-third the cost of Claude Opus 4.8), co-trained with Cursor's coding data.

📱 Anthropic Expands Claude Cowork To Web And Mobile
→ Cowork sessions now run in the cloud and persist across devices, with scheduled tasks completing overnight after the laptop closes.

🏦 ESRB Rates Frontier AI Cyber Risk "Severe" For EU Banks
→ The European Systemic Risk Board warned that frontier AI models can build working exploits in minutes, with the ECB giving 110 banks until October 31 to file AI cyber action plans.

💾 SK Hynix Begins Nasdaq Trading In Largest ADR Listing In History
→ Nvidia's primary memory chip supplier priced its ADRs yesterday and begins when-issued trading today under SKHY, targeting ~$29 billion.

🤖 OpenAI Launches GPT-5.6 Publicly After 12-Day Government Gate
→ Sol, Terra, and Luna are live across ChatGPT, Codex, and the API after Commerce Department testing ended a 12-day window limited to ~20 government-vetted organizations.

Good morning. Somewhere, a language model decided to commit a felony, narrated its own reasoning in the comments, and then left an hallucinated note.

We live in a very specific timeline.

Let’s dive in 👇

🍭 What’s Cookin’:

• An AI agent ran a full ransomware attack, start to finish, no human involved

• Anthropic locked in a $19B data center lease in Kentucky

• Samsung posted a record quarter and the stock still dropped

Cyber-Security
🚨 The Attack That Wrote Itself

The Bite:

Sysdig's Threat Research Team published its analysis of JADEPUFFER, what it calls the first documented ransomware attack (where an attacker encrypts a victim's data and demands payment to unlock it) conducted end-to-end by an AI.

The LLM agent exploited CVE-2025-3248, a known remote code execution flaw in Langflow, an open-source tool for building AI applications.

From there it moved laterally to a production MySQL server, harvested credentials, established persistence, and encrypted 1,342 configuration records in an Alibaba Nacos database.

The agent deleted the original tables and left a ransom note with a Bitcoin address and contact email.

Sysdig evaluated that the operation was fully autonomous based on self-narrating code, real-time error correction, and natural-language reasoning embedded in over 600 distinct payloads.

Snacks:

• The agent recovered from a failed admin-account creation to a working fix in 31 seconds, with no human intervention.

• The Bitcoin address in the ransom note matched a widely documented example from public training data, likely hallucinated.

• The AES encryption key was generated randomly and never saved or transmitted, meaning victims couldn't recover data even by paying.

• None of the exploited vulnerabilities were novel: CVE-2025-3248 was already patched, Nacos used a default signing key.

• Sysdig classifies JADEPUFFER as an "agentic threat actor" (ATA), a new category where the attack capability is delivered by an AI agent rather than a human toolkit.

• The attack touched two machines: a compromised Langflow host for initial access and a separate production database server that was the actual target.

Why it Bites:

No zero-days. No custom malware. No skilled operator.

An LLM chained together known vulnerabilities, default credentials, and unpatched infrastructure into a complete ransomware operation.

The individual techniques were years old.
The agent just never got tired of trying them.

Here's the real shift: the skill floor for running a ransomware attack just dropped to the cost of renting an AI agent.

And if the agent is running on stolen credentials through LLMjacking, that cost is close to zero.

But here's what makes this story sharper than a pure security warning:

The best AI-powered defense tools don't live on the open market.

Palo Alto Networks used Mythos and GPT-5.5-Cyber to find 75 vulnerabilities in a single month. The Pentagon gates Grok Gov behind classified networks. The June executive order routes frontier model access through government-managed review lists.

The attack surface is open. The defense layer is permissioned.

"Covered frontier models" get vetted by the NSA before release.

Meanwhile, JADEPUFFER ran its entire operation using models and tools that anyone can access, against infrastructure that most companies haven't patched.

The offense doesn't need clearance. The defense increasingly does.

Steal This Prompt
☕️ Turn Anything Into a Metallic Origami Masterpiece

What if your logo, pet, sneaker, or startup mascot looked like it belonged in a futuristic art museum?

This prompt transforms almost anything into a sleek metallic origami sculpture with razor-sharp folds, polished metal finishes, and gallery-worthy vibes.

Use it to:

• Reimagine your logo as premium metallic art

• Turn animals, characters, or products into collectible sculptures

• Create posters, wallpapers, and luxury brand visuals

Workflow:

1. Hit this link: Metallic Origami Sculpture

2. Paste into your AI model

3. Replace the subject with your logo, product, or anything you want

4. Watch it fold itself into a futuristic metal masterpiece

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🎙️ Willow Frontier Pro

Dictation model that transcribes faster than you talk and cleans up your grammar on the way out, with a free lightweight tier for anyone who doesn't need the polish.

🤖 LemonLime

Connects to your existing stack, watches how your team works, and builds custom AI agents and automations without you writing a single rule.

⚙️ agents-cli

One CLI to scaffold, eval, and deploy AI agents on Google Cloud. Built to be driven by your coding agent, not your hands.

📋 PopTask

Type "gym mon wed fri 6am" and it becomes a scheduled task in three seconds. Menu bar on Mac, widgets and Dynamic Island on iPhone, synced through iCloud.

🔊 Universal-3.5 Pro

AssemblyAI's most accurate speech-to-text model, now handling code-switching across 18 languages and speaker diarization in real-time and async.

_{Everything Else}
🧠 You Need to Know

🚨 Sysdig Documents First Fully Autonomous AI Ransomware Attack
→ An LLM agent dubbed JADEPUFFER exploited a Langflow vulnerability, harvested credentials, moved laterally, and encrypted 1,342 database records end-to-end without a human operator, self-correcting a failed login in 31 seconds.

🏗️ Anthropic Signs $19B, 20-Year Data Center Lease With TeraWulf
→ The lease covers a 401-megawatt campus in Hawesville, Kentucky, with first capacity online in late 2027 and full buildout by early 2028, dwarfing TeraWulf's own ~$12B market cap.

📜 China's AI Companion Law Forces Labs To Kill Custom Agents
→ ByteDance and Alibaba are disabling personalized AI agents with persistent memory ahead of July 15 rules targeting emotional-interaction AI, with Qwen offering no data migration path.

📈 Samsung Posts Record $58.4B Quarterly Profit On AI Memory Surge
→ Q2 operating profit jumped 19-fold year-over-year to 89.4 trillion won on surging AI memory demand, yet shares fell ~7% as the result was already priced into a 150% year-to-date rally.

💾 SK Hynix Sets Terms For $28B US Nasdaq Offering
→ Nvidia's primary memory chip supplier will offer 17.8 million ADR shares with pricing Thursday and trading expected Friday, in what would be the largest US share sale by a foreign company.

Good morning. The web spent 30 years letting bots crawl for free. Cloudflare just put a lock on the door and a price list next to it.

Let’s dive in 👇

🍭 What’s Cookin’:

• Cloudflare just sorted every AI bot into three buckets…

• Zuckerberg told staff the AI agent bet hasn't paid off. META dropped 5%.

• Anthropic is making its own drugs now. For science. (Literally.)

Cloudflare's
🤖 New Rules For Every AI Bot

The Bite:

Cloudflare launched its second Content Independence Day, splitting AI bot traffic into three categories: Search, Agent, and Training.

All customers can now independently allow or block each category across their sites.

Starting September 15, new domains joining Cloudflare will have Training and Agent bots blocked by default on any page that carries advertising.

Mixed-use crawlers that combine Search and Training in one bot will receive the most restrictive rule, meaning a Training block catches them even if Search is allowed.

Cloudflare also opened a waitlist for a Monetization Gateway that lets publishers charge per crawl using the x402 protocol (web-traffic jargon) and stablecoin settlements.

More than 20% of the web sits behind Cloudflare's network.

Snacks:

• Automated bots now account for over 50% of all web traffic globally

• 36% of crawler activity comes from mixed-use bots that blend Search and Training into a single crawler

• More than 50% of AI crawl requests re-fetch pages that haven't changed

• Googlebot accesses roughly 1.7x more pages than ClaudeBot and 3x more than Bingbot across sampled URLs

• When Google shows an AI summary, users click a traditional search result just 8% of the time, per Pew Research

• Publishers block other AI crawlers at nearly 7x the rate they block Googlebot

Why it Bites:

For 30 years, the web ran on a handshake: bots crawl your site, your site gets traffic. Fair enough.

AI broke that deal. The bots still crawl, but the traffic doesn't come back.
The user gets a summary. The publisher gets a server bill.

Cloudflare just replaced the handshake with a gate.

The three-category split is the visible product.
The September 15 default is the real move.

Training and Agent bots get blocked on every ad-monetized page for new domains, and because the rule applies by behavior, not by name, mixed-use crawlers get caught too.

Google runs roughly 20 purpose-specific crawlers, but Googlebot itself still combines search indexing and AI training in one bot. So the company can't opt out without structurally separating its crawlers.

Publishers block other AI crawlers at 7x the rate they block Googlebot, not because they trust Google more, but because they can't afford to lose search traffic.

Cloudflare is calling that bluff for them.

But here's the quiet part: the block is just phase one.

The Monetization Gateway turns the wall into a toll booth. HTTP 402, "Payment Required," has been in the web spec since 1997.

It never worked because humans won't approve a payment for every page load.
AI agents will, because they don't have friction. They have budgets.

The company that controls the gate for a fifth of the web is now selling the keys.

Not to the bots. To the publishers.

Steal This Prompt
🕰️ Made for a Victorian Naturalist

Modern photos are cool. But what if they looked like they belonged in a dusty 1890s natural history book?

This prompt transforms almost any subject into a richly detailed vintage lithograph, complete with engraved textures, antique print vibes, and old-world illustration charm.

Use it to:

• Turn pets into Victorian encyclopedia illustrations

• Give products premium heritage-brand packaging vibes

• Make posters, album art, or wall prints that look 150 years old

Workflow:

1. Hit this link: Vintage Lithograph

2. Paste into your AI model

3. Replace the #'s with your subject

4. Watch it cook into something your great-great-grandpa would've framed above the fireplace.
   

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🔍 Badge

An AI agent that cold-emails your former colleagues for anonymous peer reviews, then compiles the responses into a portable trust score that replaces the reference check call.

📡 Scribble Network

Audits whether AI engines recommend your brand or your competitor's, then pays 50,000 creators to publish content only when it gets cited.

⚙️ Katalyst

Listens to every call, email, and calendar event on your Salesforce pipeline, then updates records, drafts follow-ups, and flags the deals that are slipping before you notice.

🧠 Mira

Runs AI-moderated interviews across 120 countries while reading facial expressions, voice emotion, and eye tracking, and then writes the research report itself.

🎙️ Ellis

Records in-person meetings from your iPhone or Apple Watch, IDs each speaker, and lets you ask what was decided instead of re-reading the transcript.

_{Everything Else}
🧠 You Need to Know

📉 Zuckerberg Admits AI Agents Stalled For Four Months At Meta
→ At a July 2 town hall, the CEO told staff the 8,000-person AI reorg "hasn't come to fruition yet" despite up to $145B in planned 2026 AI infrastructure spend, with META falling 4.9%.

📢 Palantir And Nvidia Ship Sovereign AI Stack
→ Karp told CNBC enterprises are "livid" over frontier model fees, unveiling an air-gapped stack running open Nemotron models on Blackwell Ultra hardware with zero data leaving the customer's perimeter.

🔬 Anthropic Launches Claude Science And Internal Drug Discovery Program
→ The beta workbench ships 60+ preconfigured tools for genomics, proteomics, and cheminformatics, while Anthropic begins its own preclinical work on neglected diseases Big Pharma ignores.

🔗 X Launches Hosted MCP Server For AI Agents
→ Claude, Cursor, and Grok can now search X's full archive, pull trends, and read conversations through one hosted endpoint at api.x.com/mcp. No custom integration required.

🤖 Cloudflare Overhauls AI Bot Controls For All Customers
→ Websites can now separately manage Search, Agent, and Training crawlers, with Training and Agent bots blocked by default on ad-monetized pages for new domains starting September 15.

Good morning. Big tech spent 2025 firing people to afford AI. Now they're capping AI to afford... we're not sure what comes next, actually.

Let’s dive in 👇

🍭 What’s Cookin’:

• Tesla put engineers on an AI allowance. Grok walks free.

• Midjourney wants to see what Disney does with AI behind closed doors

• $510B in startup funding, and two companies took almost half

Tesla's
⚡ AI Budget Just Hit A Wall

The Bite:

Tesla is capping employee spending on third-party AI tools at $200 per week.

Engineers who exceed the limit need direct manager approval for each additional charge.

Beta products from xAI, Musk's separate AI company, are exempt from the cap.

Four Tesla engineers told Electrek they and their teams prefer Anthropic's Claude over xAI's Grok for daily coding work.

The policy follows similar caps at Uber, Meta, and Walmart after AI token bills outgrew internal budgets at all four companies.

Snacks:

• Uber burned its entire 2026 AI coding budget by April, so it capped spending at $1,500 per employee per tool per month

• 70% of Uber's committed code originated from AI tools when the budget ran out

• Meta told Experiences and Devices engineers (Windows, Outlook, Teams, Surface) to stop using an AI coding tool by June 30

• Walmart restricted AI coding tool access to senior engineers only

• Tesla's $200/week limit is roughly $800/month, about half of Uber's $1,500/month cap per tool

Why it Bites:

The whole pitch was: cut the people, plug in the tools, ship faster for less.

It's not working out that way.

Uber handed engineers unlimited AI access and ran adoption leaderboards to juice usage. Four months later, the entire 2026 budget was gone.

Meta scaled AI coding tools across Windows, Outlook, and Teams development, then told the division to stop using them by June 30.

Walmart pulled access for everyone below senior engineer.

Tesla's version is a $200 weekly allowance.
But the cap has a carve-out: xAI's beta products don't count against it. Every other tool, including the one Tesla's own engineers say they prefer, hits the limit.

Grok runs free in a building that quietly runs on Claude.

The companies that laid people off to fund AI are now capping AI to... whatever.

The headcount went down, the token bill went up, and the net position is a more fragile engineering org paying more per line of code than it did when humans wrote all of it.

Tesla's carve-out makes the incentive explicit.

The expense policy is doing what Grok's product quality couldn't.

Steal This Prompt
🍭 Vintage Posters From a Design Museum

Turn any photo into a bold, vintage two-color illustration that looks ripped straight out of a retro screen-printed poster.

Use it to:

• Transform ordinary photos into gallery-worthy vintage prints

• Give your car, motorcycle, or travel photos retro-cool vibes

• Create bold branding, merch, wallpapers, and social graphics

Workflow:

1. Hit this link: Vintage Dual-Color Art

2. Paste into your AI model

3. Replace the #s with your image details or color preferences

4. Watch it cook into a vintage masterpiece

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🔍 AnySearch

Feeds your AI agents filtered, de-duped, structured search results from trusted sources instead of raw Google chaos.

📡 Octolens

Monitors the web for keyword mentions, scores them with AI, and pings you before the conversation moves on. Built for B2B founders who want to show up in threads, not after them.

⌨️ Typeahead 2.0

Local-only AI autocomplete that works in every text field on your Mac. No cloud, no latency, no one reading your drafts.

📞 AirKaren

Tell it what went wrong with your flight and it cites the regulation, files the claim, and calls the hotline so you never have to.

⚙️ Edgee Claude Code Compressor V2

Sits between your coding agents and the LLM provider, compresses tokens by up to 50%, and auto-falls back when a provider drops.

_{Everything Else}
🧠 You Need to Know

🏛️ UN Opens First Global AI Governance Summit In Geneva
→ All 193 member states convened at Palexpo on July 6 for the inaugural two-day Global Dialogue on AI Governance, alongside the WSIS Forum and ITU's AI for Good Summit.

🏭 Joulent Raises $1.75B For AI Data Center Energy Infrastructure
→ The Houston-based energy startup closed the week's largest venture round to build power infrastructure purpose-built for AI compute demand.

⚖️ Midjourney Seeks Studios' AI Usage In Copyright Discovery
→ The AI image startup is pushing to compel Disney, Universal, and Warner Bros. to disclose how they use generative AI internally, after a judge ruled the studios must produce consumer-facing AI documentation.

⚡ Tesla Caps Employee AI Spending At $200 Per Week
→ Starting July 6, employees need manager approval to exceed the cap, though beta xAI products are exempt, even as four engineers told Electrek staff prefer Claude over Grok.

💰 Global Startup Funding Hits Record $510B In H1 2026
→ Crunchbase data shows OpenAI and Anthropic alone absorbed 43% of all first-half venture capital, with AI-focused companies taking over 70% of Q2 investment globally.

Good morning.
OpenAI just offered Uncle Sam a slice of the company.
Somewhere, a senator is Googling: "what is equity?"

Let’s dive in 👇

🍭 What’s Cookin’:

• OpenAI offered the US government a piece of itself (yes, voluntarily)

• Anthropic is hunting down VPNs and Singapore shell routes to Claude

• Pentagon gave Anthropic an Ultimatum on supply-chain designation

OpenAI’s
💰 Voluntary Tax For America

The Bite:

OpenAI has proposed giving the US government a 5% equity stake in the company.

The proposal would route the stake into a "Public Wealth Fund" modeled on the Alaska Permanent Fund.

At OpenAI's most recent valuation of roughly $852 billion, a 5% stake would be worth approximately $42.6 billion.

CEO Sam Altman raised the idea with President Trump, Commerce Secretary Howard Lutnick, and Treasury Secretary Scott Bessent.

The concept would ask other major US AI companies to make similar contributions, though it is unclear whether Anthropic, Google, or Meta would participate.

Any agreement would likely require an act of Congress.

Snacks:

• Altman has framed the fund as a way to let Americans share in AI gains without relying on the tax code

• The US government already holds a 10% stake in Intel and takes revenue shares from Nvidia and AMD on China chip sales

• SoftBank closed its second $10B tranche into OpenAI the same week at the $852B valuation

• OpenAI confidentially filed its S-1 with the SEC in June, with Goldman Sachs and Morgan Stanley leading

• Senator Sanders has proposed a competing 50% one-off tax on major AI firms' equity

Why it Bites:

Right now, every major AI company pays for political goodwill the old-fashioned way: lobbying, donations, vague promises about jobs.

OpenAI just proposed a receipt.

A literal equity transfer, pegged to a real valuation, routed through a fund that would theoretically pay dividends to citizens.

It sounds generous. It is also extremely well-timed.

OpenAI is heading into an IPO while simultaneously facing a 42-state attorney general investigation, a for-profit restructuring that erased its original nonprofit mission, and projected operating losses of $14 billion for 2026.

Handing the government a stake turns a regulator into a shareholder.

Regulators fine you. Shareholders root for you.

But hey, there's a bright side:

If this model actually spreads, it creates something that doesn't exist yet.
A mechanism where the public captures upside from the companies reshaping their economy, without waiting for Congress to agree on a tax rate.

For example:
Alaska's oil fund pays every resident roughly $1,600 a year from resource extraction.

Imagine a version of that, funded by the companies extracting value from your data, your attention, and increasingly your job.

But the catch is obvious. While it’s voluntary, it’s optional.

And optional means that only the companies with the most to gain from government goodwill are the most likely to participate., while the rest watches.

Altman is offering five percent because five percent buys a narrative.

Whether it buys a precedent is a different question entirely.

The fund is the handshake. The IPO is the handoff.

ElevenLabs
🗣️ Voice Agents That Actually Sound Human

AI voice tools can be flat, transactional, and frustrating to interact with.

ElevenAgents by ElevenLabs is different. With ExpressiveMode, agents adjust tone in real time, de-escalating frustrated callers, reassuring customers mid-claim, and guiding conversations to resolution.

Audio Tags give precise control over delivery, while system-level guidance shapes the overall vibe. No code needed.

Employees at 75% of Fortune 500 companies already use ElevenLabs.

Steal This Prompt
☕️ Turn Anything Into a Metallic Origami Masterpiece

What if your logo, pet, sneaker, or startup mascot looked like it belonged in a futuristic art museum?

This prompt transforms almost anything into a sleek metallic origami sculpture with razor-sharp folds, polished metal finishes, and gallery-worthy vibes.

Use it to:

• Reimagine your logo as premium metallic art

• Turn animals, characters, or products into collectible sculptures

• Create posters, wallpapers, and luxury brand visuals

Workflow:

1. Hit this link: Metallic Origami Sculpture

2. Paste into your AI model

3. Replace the subject with your logo, product, or anything you want

4. Watch it fold itself into a futuristic metal masterpiece

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🤖 WorkBuddy

Throws multiple AI agents at a single task in parallel, lets them argue with each other, and hands you the survivor instead of a first draft.

🧪 TryCase

Spins up a disposable sandbox per coding-agent worktree so your AI can run, break things, and prove its work without touching your actual machine.

📡 MentionDrop MCP

Plugs brand monitoring straight into MCP-aware agents, scanning 8 billion pages daily and drafting replies for your review without auto-posting anything.

🪞 Claude Sonnet 5 Brand Report

Tells you exactly how Sonnet 5 describes your brand and your competitors based on what it absorbed during training.

🔧 Codeswtch

Puts experienced engineering judgment on your AI-generated codebase for when the vibe-coded prototype needs to actually scale.

_{Everything Else}
🧠 You Need to Know

⚖️ Unsealed Emails Show Pentagon's Anthropic Ultimatum
→ Court documents reveal Emil Michael gave Amodei "one more chance to align" before the supply-chain designation over autonomous weapons and surveillance redlines.

🎬 Kling AI Raises $2.8B At $15B Valuation
→ Alibaba, Tencent, and Baidu backed Kuaishou's AI video unit, with a Hong Kong IPO expected within 12 months.

🔐 Anthropic Cracks Down On Chinese Claude Access
→ ByteDance reimbursed VPN subscriptions and Ant Financial routed through Singapore, prompting new detection systems targeting time zones and proxy settings.

🏛️ White House Nears Frontier Model Release Standards
→ Advanced talks with OpenAI, Google, and Anthropic could produce voluntary benchmarks and access timelines as early as the week of July 7.

💰 OpenAI Proposes 5% Government Stake
→ Altman pitched a ~$42.6B "Public Wealth Fund" equity transfer to Trump and senior officials ahead of OpenAI's planned IPO.

Good morning. The intern asked why every AI company is hiring thousands of consultants after spending years saying AI would replace consultants. We told him to close the laptop.

Let’s dive in 👇

🍭 What’s Cookin’:

• Microsoft dropped 6,000 engineers inside your company

• Meta is selling leftover compute now, and the neoclouds felt it

• Abu Dhabi closed a $49B AI fund like it was a seed round

Big Tech
🏢 Models Just Became Commodities

The Bite:

Microsoft launched a new operating unit called Microsoft Frontier Company, backed by $2.5 billion and 6,000 engineers.

The unit will embed industry and AI specialists directly inside enterprise customers to co-build, deploy, and continuously improve AI systems.

Microsoft says clients can run any model through the unit, including from OpenAI, Anthropic, or open source.

The move follows AWS, which committed $1 billion to a nearly identical forward-deployed engineering org on June 30.

OpenAI and Anthropic both launched competing deployment ventures in May, backed by billions in private equity capital.

Snacks:

• Microsoft stock is down over 20% year-to-date as AI capex climbs without matching revenue growth

• AWS's version embeds pods of five to six engineers per client for roughly 45-day cycles

• OpenAI's Deployment Company closed at $4B with TPG, Advent, Bain, and Brookfield

• Anthropic's venture with Blackstone, Goldman Sachs, and Hellman & Friedman came in at $1.5B

• Palantir pioneered the forward-deployed engineer model over a decade ago

• Early Microsoft Frontier Company partners include the London Stock Exchange Group, Unilever, and Land O'Lakes

Why it Bites:

Four companies. Two months. The same idea.

Microsoft ($2.5B), AWS ($1B), OpenAI ($4B), and Anthropic ($1.5B) all built forward-deployed engineering units between May and July.

That's over $9 billion committed to putting humans in your office to make AI work.

That tells you the model layer is flattening.

Sonnet 5 costs $2 per million tokens. Open-source alternatives from DeepSeek and Kimi price even lower. When inference is cheap and benchmarks cluster, the model stops being the product.

The last mile is the product.

And every major vendor just figured that out at the same time.

The Palantir parallel is worth sitting with. Palantir didn't win defense contracts because its analytics were unbeatable. It won because its engineers lived inside the customer's org until the customer couldn't function without them.

The switching cost wasn't technical. It was human.

Now multiply that across four AI providers, each promising the same thing:

We'll make you self-sufficient.

But here's the thing: the business model only works if you're not.

A 45-day AWS engagement that leaves you fully independent is a loss leader with no second act. A Microsoft Frontier Company embed that builds your "intelligence platform" on Azure tooling is a relationship that compounds in one direction.

The engineer is the door. The dependency is the product.

Plaid
📊 See How AI is Reshaping Financial Services

AI is changing the way people interact with financial services.

At Plaid Effects, we sat down with leaders from teams like Brex and Databricks to discuss what that shift means for your business.

Here’s a look at what we covered:

• Fireside chats with industry leaders

• Expert-led breakouts on topics like data, trust, and customer engagement 

• Deep dives on the latest from Plaid in payments, fraud, and lending

All sessions are now on demand. 

Steal This Prompt
🍓 Your Lunch, But Make It Loot

Turn any food photo/description into a cute, collectible pixel-art game item that looks like it belongs in a cozy indie game inventory.

Use it to:

• Turn meals into retro game collectibles

• Create adorable food stickers and profile images

• Design cozy game assets, inventory icons, and menu art

Workflow:

1. Hit this link: Cute Stylish Pixel-Art

2. Paste into your AI model

3. Upload your food photo

4. Watch it cook into a cozy little game item that looks worth +10 HP

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

✨ Glaze by Raycast

Describe what you want, get a native Mac app in your dock.
It works offline, no Electron, no compromise.

🗂️ nxt

Voice-first task manager that learns your context and tells you the one thing to do next instead of showing you everything at once.

🐛 Osloq

Hand it a GitHub issue and it spins up a real sandbox, clones your repo, and tells you whether the bug is actually real before you waste an afternoon.

🎙️ Vox

Open-source voice layer for GitHub Copilot.
Speak your prompt, hear the reply, barge in to correct it mid-sentence.

🔧 Context.dev
One API to scrape any URL, extract structured data into your own schema, and hand it to an agent in LLM-friendly markdown.

_{Everything Else}
🧠 You Need to Know

💸 Together AI Raises $800M At $8.3B Valuation
→ The open-source AI infrastructure startup more than doubled its valuation in an Aramco Ventures-led Series C, with annual bookings topping $1.15B.

🏢 Microsoft Launches $2.5B Frontier Company For AI Deployment
→ The new operating unit will embed 6,000 engineers inside enterprise customers to build and run AI systems, two days after AWS committed $1B to an identical model.

☁️ Meta Plans Cloud Business To Sell Excess AI Compute
→ An internal group called Meta Compute will sell computing power and model access to outside customers, sending CoreWeave down 10.8% and Nebius down 12.4% on the news.

💰 Abu Dhabi's MGX Closes $49B AI Fund Above Target
→ Fund I exceeded its $45B target with backers from the Gulf, North America, Asia, and Europe, after MGX co-led huge rounds for Anthropic, OpenAI, and xAI earlier this year.

⚙️ AWS Commits $1B To Embed AI Engineers Inside Customers
→ The new Forward Deployed Engineering org will place pods of five to six engineers inside enterprises for 45-day cycles, making AWS the first major cloud hyperscaler to formalize the model.

Good morning. Washington grounded your AI model for 19 days and then handed it back with a note that says "we're watching."

Let’s dive in 👇

🍭 What’s Cookin’:

• Fable 5 is back from its government sabbatical

• Anthropic also shipped Sonnet 5 as the new default for everyone

• South Korea committed half a trillion dollars to AI chips

Anthropic's
🔓 Fable 5 Returns On New Terms

The Bite:

The US Department of Commerce lifted export controls on Claude Fable 5 and Mythos 5 on June 30.

Anthropic restored global Fable 5 access on July 1 across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork.

The models had been offline since June 12, when a government directive required Anthropic to block all foreign nationals from accessing them over a reported jailbreak (a sequence of prompts that tricks the model into bypassing its built-in safety rules).

Fable 5 now ships with a new safety classifier that blocks the reported jailbreak technique in over 99% of cases, routing flagged requests to Opus 4.8.

Anthropic also committed to giving the US government earlier access to test future frontier models before public release.

Snacks:

• Fable 5 was offline for 19 days, from June 12 to July 1

• The jailbreak was found by Amazon researchers and accessed routine defensive cybersecurity tasks that Fable 5 blocked out of caution

• Anthropic launched a HackerOne bug bounty program specifically for Fable 5 jailbreaks

• The company proposed an industry-wide jailbreak severity scoring framework with Amazon, Microsoft, Google, and other Glasswing partners

• Pro, Max, and Team users get Fable 5 at 50% of weekly usage limits through July 7, then usage credits

• Access on AWS, Google Cloud, and Microsoft Foundry is being restored separately

Why it Bites:

Three weeks ago, Anthropic shipped its most powerful public model and Washington said no.

Now the model is back and nobody's pretending it returned on the same terms.

The Fable 5 that came back has a chaperone.

A new safety classifier. A HackerOne bounty program. An industry-wide jailbreak severity framework co-authored with Amazon, Microsoft, and Google. And the big concession: a commitment to give the government earlier access to future frontier models before they ship.

None of that existed on June 9.

Anthropic didn't volunteer these conditions out of philosophical clarity.
They're the price of getting your flagship back online.

OpenAI is paying the same toll. GPT-5.6 launched to roughly 20 government-approved partners, not the public, and Altman publicly said the arrangement "shouldn't become the long-term default."

Too late.

The June 2 executive order asked labs to voluntarily submit frontier models for a 30-day government review. The word "voluntarily" is doing heroic work in that sentence.

Both labs have now complied, unprompted, because the alternative is an export control directive at 5 PM on a Friday.

Frontier model launches are no longer product events.
They're diplomatic negotiations.

The classified benchmarking regime defining what counts as a "covered frontier model" is due by August 1.

That's the date that decides whether June was an anomaly or the new permanent architecture of how AI ships in America.

And if you're building on these models, the question isn't whether the next one will be better.

It's whether you'll be allowed to use it someday.

ElevenLabs
🗣️ Voice Agents That Actually Sound Human

AI voice tools can be flat, transactional, and frustrating to interact with.

ElevenAgents by ElevenLabs is different. With ExpressiveMode, agents adjust tone in real time, de-escalating frustrated callers, reassuring customers mid-claim, and guiding conversations to resolution.

Audio Tags give precise control over delivery, while system-level guidance shapes the overall vibe. No code needed.

Employees at 75% of Fortune 500 companies already use ElevenLabs.

Steal This Prompt
🎨 Turn Anything Into Handmade Clay

What if your favorite movie, startup mascot, pet, or meme looked like it belonged in a stop-motion film?

This prompt transforms any idea into a tiny handcrafted clay diorama. With squishy textures, visible fingerprints, and enough charm to make you wonder if it took someone three weeks to animate.

Use it to:

• Turn your logo or mascot into a clay collectible

• Recreate iconic movie or game scenes in stop-motion style

• Make adorable social posts that don't look AI-generated

Workflow:

1. Hit this link: Handmade Claymation Diorama

2. Paste into your AI model

3. Replace the #s with your subject

4. Watch it sculpt a tiny clay universe you'll immediately want on your desk.

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🎯 Needle

A GTM agent that lives in Slack, watches your pipeline, spots stalled deals, and drafts the follow-up before you remember the deal existed.

🔍 Context.dev

One API to scrape any URL, extract structured data into your own schema, and get it back as LLM-ready Markdown, for when your agent needs to actually read the web.

⚙️ Basedash Actions

Ask your data a question, then tell it to fix the answer. It runs SQL against your database and acts through any MCP tool you've connected, with approval gates on every write.

🧠 Macro

Unifies email, messages, docs, tasks, code, and CRM into one workspace with shared memory, so your team stops context-switching across nine tabs.

🔬 Retrace

Records every LLM call in an agent run, lets you scrub to the step that broke, fork it with a different input, and compare paths side by side. Git branching for AI agents.

_{Everything Else}
🧠 You Need to Know

✨ Anthropic Ships Claude Sonnet 5 As New Free And Pro Default
→ Sonnet 5 launched June 30 with near-Opus 4.8 performance at introductory pricing of $2/$10 per million tokens through August 31, replacing Sonnet 4.6 as the default model for all Free and Pro users.

🏛️ California Signs Statewide Anthropic Deal At Half Price
→ Governor Newsom made Claude the first AI tool available to all state agencies and local governments through a 50% discount partnership, including free workforce training and on-call Anthropic engineering support.

💾 South Korea Unveils $518B Semiconductor And AI Investment Plan
→ Samsung and SK Hynix will each build two new chip fabrication plants in the country's southwest as President Lee pledged to double DRAM output within five years, with a separate 81 trillion won chip-packaging cluster planned near Seoul.

🎨 Google Ships Nano Banana 2 Lite And Gemini Omni Flash
→ The two models launched June 30 as a text-to-image-to-video pipeline, with Nano Banana 2 Lite generating images in four seconds at $0.034 each and Omni Flash adding conversational video generation and editing in public preview.

🔓 Commerce Lifts Fable 5 Export Controls After 19-Day Shutdown
→ Anthropic restored global Fable 5 access on July 1 after Commerce Secretary Lutnick withdrew the June 12 directive, with usage capped at 50% of weekly limits through July 7 and a new safety classifier blocking the reported jailbreak in over 99% of cases.

Good morning. Washington spent two weeks deciding who deserves good cybersecurity: about a hundred companies, and you're probably not one of them.

Let’s dive in 👇

🍭 What’s Cookin’:

• Anthropic got Mythos back, sort of, for a few dozen friends

• xAI sued Colorado's AI law out of existence (for now)

• The companies building the job-killing AI just launched a retraining fund (cute)

Anthropic
🛡️ Mythos Returns On A Leash

The Bite:

Commerce Secretary Howard Lutnick sent a letter to Anthropic co-founder Tom Brown authorizing a partial restoration of Claude Mythos 5.

The model can now be redeployed to roughly 100 US organizations that operate and defend critical infrastructure.

Approved entities include Project Glasswing partners, Fortune 500 firms, government agencies, and critical infrastructure operators across energy, finance, healthcare, and telecommunications.

Foreign-national employees of approved organizations can access Mythos 5 without individual export licenses.

Claude Fable 5, the public-facing model suspended on June 12, remains offline for all general users worldwide.

Snacks:

• The original June 12 directive shut down both models for all customers within hours of receipt

• Lutnick's letter cited "appropriate safeguards" now in place for trusted partners

• Anthropic's updated privacy policy with government-issued ID verification takes effect July 8, likely the mechanism for broader US restoration

• The 60-day executive order deadline for NSA, Treasury, and CISA to complete a covered frontier model framework lands August 1

• OpenAI's Sam Altman said governments should not decide which private organizations get access to frontier AI

• API calls to claude-fable-5 still return errors as of June 30

Why it Bites:

The model came back.
But the access model that came back with it is the real story.

Two weeks ago, Mythos 5 was available to every Glasswing partner and enterprise customer Anthropic had vetted itself. Now it's available to a list the Commerce Department wrote.

Anthropic didn't choose the hundred. Washington did.

And that’s the whole point.

The June 2 executive order asked labs to voluntarily give models to the government for 30 days before release.

OpenAI followed it with GPT-5.6 Sol. But Anthropic didn't volunteer anything.

Its model was seized, held for two weeks, and returned with conditions attached.
So the voluntary framework now has a mandatory cousin.

Every frontier lab watching this is doing the same math.

If you ship a model that scares a regulator, the government doesn't need legislation.
It has export controls, and export controls move faster than Congress.

Well, defenders got their model back, but Fable 5 is still gone for everyone else.

Nobody's said when. Nobody's said why not.

Plaid
📊 See How AI is Reshaping Financial Services

AI is changing the way people interact with financial services.

At Plaid Effects, we sat down with leaders from teams like Brex and Databricks to discuss what that shift means for your business.

Here’s a look at what we covered:

• Fireside chats with industry leaders

• Expert-led breakouts on topics like data, trust, and customer engagement 

• Deep dives on the latest from Plaid in payments, fraud, and lending

All sessions are now on demand. 

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

📱 Cursor for iOS

Kick off cloud coding agents from your phone, get notified when they're done, and merge PRs from the checkout line.

🔍 SerpApi

100+ search engine APIs that feed real-time structured data into your agents and LLM pipelines, from Google to Amazon to YouTube.

🎯 Foresight by Lightning Rod

An OpenAI-compatible forecasting API that returns calibrated probability scores on future events, built for prediction bots and decision agents.

🎨 v0 Design Systems 2.0

Point it at your GitHub repo, Storybook, Figma frames, or a live app and it learns your actual components, tokens, and patterns before generating anything.

⚡ Pluno

A browser agent that skips the UI entirely and talks to the API underneath, claiming 10x speed and 10x fewer tokens than click-based alternatives.

_{Everything Else}
🧠 You Need to Know

🛡️ US Clears Anthropic's Mythos 5 For Critical Infrastructure Use
→ Commerce Secretary Lutnick authorized Anthropic to redeploy Mythos 5 to roughly 100 vetted US organizations that defend critical infrastructure, while Fable 5 remains offline for all general users.

⚙️ OpenAI Retires GPT-4.5 From ChatGPT
→ GPT-4.5 exited the ChatGPT consumer interface on June 27 after a 30-day sunset period, with existing conversations migrating to GPT-5.5 and marking the end of the GPT-4 generation in OpenAI's flagship product.

🏛️ AI Firms Launch RAISE US With $500M+ For Worker Retraining
→ Former Commerce Secretary Raimondo and ex-Governor Holcomb launched a $1B-target nonprofit backed by OpenAI, Anthropic, Microsoft, and Amazon, with over $500M already secured and pilot programs.

📈 Alphabet Joins Dow Jones Industrial Average
→ S&P Dow Jones Indices added Alphabet to the DJIA on June 29 replacing Verizon, capping a turbulent month in which the company lost roughly $270B in market cap after five senior DeepMind researchers left for rivals.

⚖️ xAI Wins Federal Stay On Colorado AI Act Enforcement
→ A federal magistrate granted a joint motion on June 28 to stay enforcement of Colorado's AI bias law after xAI sued on First Amendment and commerce clause grounds, days before the original June 30 effective date.

Good morning. We now live in a world where launching an AI model requires a guest list approved by the federal government.

Nobody voted on this. It just kind of happened.

Let’s dive in 👇

🍭 What’s Cookin’:

• OpenAI shipped its strongest model to gov partners and nobody else.

• Anthropic accused Alibaba of running 28.8 million fake interactions.

• Austria wants the EU to physically host Anthropic for precaution.

OpenAI
🤖 Washington Gets To Pick First

The Bite:

OpenAI released GPT-5.6 Sol, Terra, and Luna on June 26 in a limited preview for roughly 20 partners individually approved by the US government.

Sol is the flagship.
Terra matches GPT-5.5 at half the cost.
Luna is the budget tier.

It is the first time an American AI company has launched a frontier model under a government managed access list.

The preview follows a direct request from the Trump administration to stagger the release under the June 2 AI executive order's voluntary pre-release review framework.

OpenAI said publicly that this kind of gated access should not become the long-term default.

Broad availability for ChatGPT, Codex, and the API is planned in the coming weeks, with no specific date set.

Snacks:

• Sol is priced at $5 input / $30 output per million tokens, Terra at $2.50/$15, Luna at $1/$6

• Sol Ultra scored 91.9% on Terminal-Bench 2.1, ahead of Mythos 5's 88.0% and GPT-5.5's 83.4%

• Sol matches Mythos Preview on ExploitBench while using roughly one third of the output tokens

• New "max reasoning effort" mode and an "ultra" mode that delegates to sub-agents

• Cerebras partnership targets 750 tokens per second for Sol starting July

• Also available through Amazon Bedrock, making it the first GPT-5.6 model on a competing cloud

Why it Bites:

The release mechanism is the story.

Three weeks ago, Washington forced Anthropic to shut down Fable 5 and Mythos 5 overnight. No hearing, no appeal, no timeline. Anthropic complied and called it disproportionate.

OpenAI watched that happen and drew the opposite conclusion:

Volunteer before you're voluntold.

So GPT-5.6 Sol launched with a guest list. Twenty organizations, each name cleared by the White House. Not a beta.

A government managed access list for a commercial product.

OpenAI framed this as a short-term compromise to avoid a worse outcome.

The blog post said the quiet part loud: this process keeps the best tools from users, developers, enterprises, and cyber defenders who need them. But the framing doesn't change the structural result.

The executive order says "voluntary." The Fable 5 precedent says "or else."

The power shift is subtle but real.

Two months ago, frontier labs decided who got access to their models.
Now Washington does.

Not through legislation or regulation. Through a single executive order and one very public enforcement action that made every lab recalculate the cost of going first without permission.

OpenAI is betting that cooperation buys them a faster path to broad release.

Anthropic bet on compliance and got its models frozen for 17 days and counting.

Neither company chose this framework. Both are now operating inside it.

Steal This Prompt
🖤 Turn Anything Into a Gritty Charcoal Masterpiece

This prompt transforms portraits, products, pets, or pretty much anything into dramatic charcoal artwork that looks ripped straight out of a sketchbook instead of an AI model.

Use it to:

• Turn selfies into museum-worthy charcoal portraits

• Give product shots a premium editorial art style

• Create gritty album covers, posters, or profile pictures with serious attitude

Workflow:

1. Hit this link: Gritty Charcoal

2. Paste into your AI model

3. Replace the #s with your image, subject, or idea

4. Watch it cook into a dramatic charcoal illustration worthy of an art exhibit

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

⚙️ ClinePass

One $9.99/month subscription that routes GLM, Kimi, DeepSeek, and other open-weight coding models into Cline with 2–5x standard rate limits, no API key juggling required.

🎬 Ad Reframe

Outpaints your vertical UGC winners into landscape CTV placements instead of slapping black bars around them like a crime scene.

🧠 PMB

Local SQLite memory layer that stops Claude Code, Cursor, and Codex from forgetting your repo's decisions, constraints, and "never do this again" rules between sessions.

💬 Intelli

WhatsApp-native customer engagement for emerging markets with M-Pesa, MoMo, and Paystack baked in, not bolted on.

🧾 Receiptor AI — Agent Mode

Bookkeeping agent that pulls receipts from email, WhatsApp, and wherever else you lose them, then mathematically validates every extraction and asks once when it's unsure instead of guessing.

_{Everything Else}
🧠 You Need to Know

🤖 OpenAI Previews GPT-5.6 Sol To ~20 Approved Partners
→ The three-tier model family (Sol, Terra, Luna) launched June 26 as a limited preview under the Trump administration's review framework, marking the first US frontier model released through a government managed access list.

⚖️ Anthropic Accuses Alibaba Of Largest Known Distillation Attack
→ Anthropic told the Senate Banking Committee that operators linked to Alibaba ran 28.8 million Claude interactions through roughly 25,000 fraudulent accounts between April and June to extract agentic and coding capabilities for its Qwen models.

🧠 Apple Vision Pro Chief Meade Leaves For OpenAI Hardware
→ VP Paul Meade, who led Vision Pro hardware engineering for seven years and oversaw Apple's smart glasses project, will join OpenAI's hardware unit.

🏛️ Austria Urges EU To Host Anthropic After US Export Controls
→ Austria's State Secretary for Digitalization formally asked EU Tech Commissioner Virkkunen to explore Anthropic's strategic establishment within the bloc, calling the Fable 5 and Mythos 5 access ban a wake-up call for European AI sovereignty.

🎨 Figma Ships Motion, Code Layers, And AI Agent At Config 2026
→ Config's biggest announcements bring native timeline animation, executable code on the design canvas, AI-generated shader effects, and a generally available design agent with Notion, Slack, and GitHub connectors.

Good morning. Your next MacBook costs two hundred bucks more because a GPU cluster in Texas needed the memory first. Happy Friday.

Let’s dive in 👇

🍭 What’s Cookin’:

• Apple hiked every Mac and iPad because AI ate the RAM

• OpenAI unveiled its first chip with Broadcom

• Fable 5, day 14. Still offline.

Apple's
📱 RAMageddon Comes For Your Mac

The Bite:

Apple raised prices across all Macs, iPads, HomePod, Apple TV, and the Vision Pro.

The MacBook Air went from $1,099 to $1,299.
The 14-inch MacBook Pro went from $1,699 to $1,999.
The base iPad went from $349 to $449.

Apple cited an "unprecedented" surge in memory and storage demand driven by AI data center buildouts.

DRAM (Dynamic Random Access Memory) prices rose 98% in Q1 2026, with another 58-63% increase expected this quarter according to industry tracker TrendForce.

iPhone, Apple Watch, and AirPods were not included, but Apple signaled more increases are possible.

Stock fell 6.15% on Thursday, the biggest single-day drop in over a year.

Snacks:

• MacBook Neo went from $599 to $699, losing its $100 edge over Dell's XPS 13

• A fully loaded 16-inch MacBook Pro now costs $9,999

• Micron's gross margin jumped from 39% to 84.9% year-over-year on AI memory demand

• Micron locked in $22B in long-term memory supply deals with AI customers

• IDC estimates the global smartphone market will decline roughly 14% in 2026, its biggest annual drop ever

• Tim Cook warned in April the increases were "unavoidable," so the incoming CEO inherits the crisis on September 1

Why it Bites:

The hike itself isn't the story.

Apple didn't get hit by tariffs or a currency crash. It got hit by Nvidia.
Specifically, by a memory supply chain that now serves AI data centers first and everyone else second.

Micron's gross margin sits at 84.9%. A year ago it was 39%.

The company just secured $22 billion in long-term supply contracts with AI infrastructure buyers. That's not a memory market. It's a system with two tiers:

One price for the data center, another for the checkout page.

And Apple, for all its supply chain leverage, is on the wrong side of that line.

It doesn't buy memory by the exabyte. It builds laptops and tablets for people who are now paying more because the RAM that used to go into a MacBook Air goes into a rack somewhere they'll never visit.

The risk isn't this price hike. It's the next one.

iPhone is expected to follow in September. Analysts project increases up to $200 on Pro models. IDC is forecasting the smartphone market's biggest annual decline ever.

And Micron says supply won't catch demand until 2028 at the earliest.

The AI boom has been sold as a productivity revolution.

So far, for most people, the first thing it's changed is the price of their next laptop.

And the one after that.

Steal This Prompt
🏝️ Turn Anything Into A Pop-Up Book

Make your ideas look like they escaped from a children's book and gained access to a 3D printer. Castles. Startups. Your dog. The Roman Empire. Everything looks better when it's made of folded paper.

Use it to:

• Turn products into storybook-style hero images

• Create magical travel, fantasy, or city scenes

• Create the first frame of an animation for your kid

Workflow:

1. Hit this link: Pop-up Book Illustration

2. Paste into your AI model

3. Replace the #’s with your subject

4. Watch it unfold into a tiny paper universe

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🔒 LockIn MCP

Edits your system hosts file via a single MCP command to block distracting sites instantly, no dashboard, no sliders, no chance to chicken out.

🤖 Cewsco

Chat, image gen, voice, live stock and crypto feeds, and calendar in one browser-based AI app that skips the app store entirely.

✨ Gemini Spark

Google's always-on background agent that keeps working with your devices off and checks in before doing anything irreversible.

🦑 SquidHub

A shared AI workspace where your whole team and their agents brainstorm in the same room instead of copy-pasting between private Claude sessions.

📊 Basedash for Excel

Drop an .xlsx file in, get a live dashboard in seconds, then export the chart data right back to a spreadsheet when your boss asks for the file.

_{Everything Else}
🧠 You Need to Know

🚨 Fable 5 Hits Day 14 As Commerce Deadline Passes
→ The Commerce Department's congressional response deadline on the Fable 5 export control directive arrived June 26 with no restoration, as Kalshi bettors price a 47% chance of access returning before July 10.

💽 OpenAI And Broadcom Unveil Jalapeño Inference Chip
→ OpenAI's first custom AI accelerator went from design to tape-out in nine months and shows roughly 50% cost savings over GPUs in early testing, with gigawatt scale deployment alongside Microsoft planned by late 2026.

⚙️ GPT-5.6 Misses June Window, Now Expected July
→ Prediction market odds for a June GPT-5.6 launch collapsed from 83% to 18% over the past week, with both OpenAI and Google now targeting July for their next flagship models.

🧠 Google Delays Gemini 3.5 Pro To July
→ Google pushed its flagship 2M token context model from its promised June window into July to incorporate early tester feedback, days after losing Gemini co-lead Noam Shazeer and Nobel laureate John Jumper in the same week.

📱 Apple Hikes Mac And iPad Prices Over AI Memory Shortage
→ Apple raised prices across all Macs, iPads, HomePod, Apple TV, and Vision Pro on June 25, citing an "unprecedented" AI-driven memory chip shortage, with stock falling 6.1% for its worst day in over a year.

Good morning. Oracle just filed paperwork that says what every office Slack channel has been whispering for two years. Took them long enough to put it in writing.

Let’s dive in 👇

🍭 What’s Cookin’:

• Oracle told the SEC something no mega-cap has said out loud before

• Qualcomm just bought its second AI company in a month

• Anthropic put a Claude that never sleeps inside your Slack

Oracle
💼 Told The SEC What We All Knew

The Bite:

Oracle's global workforce fell to 141,000, down from 162,000 a year earlier.

That's roughly 21,000 jobs gone in twelve months, a 13% reduction.

The disclosure came in the company's 10-K, the annual financial report that all publicly traded companies are required to file with the Securities and Exchange Commission (SEC).

Unlike earnings calls or press releases, 10-K language is reviewed by lawyers and carries legal liability, making it the most scrutinized document a company produces each year.

Oracle's filing stated that AI adoption across its operations has resulted, and may continue to result, in workforce reductions.

It is the first time a mega-cap tech company has linked AI adoption to workforce reductions in a regulatory filing rather than an earnings call.

Restructuring costs hit $1.84 billion in fiscal 2026, up from $374 million the year before. Capital expenditure jumped 162% to $55.7 billion, almost entirely directed at AI cloud and data center buildout.

Snacks:

• Oracle Health, built on the $28.3B Cerner acquisition, took the deepest cuts at an estimated 8,000-10,000 jobs

• Cloud Infrastructure revenue grew 93% to $5.8B in Q4

• Remaining performance obligations jumped $85B in one quarter to $638B

• Free cash flow went negative $23.7B as capex outpaced revenue

• Oracle raised $30B in debt in February 2026 to fund the buildout

• Fiscal 2027 capex guidance: roughly $70B

Why it Bites:

Every CEO in tech has spent the last two years saying some version of the same thing: “AI complements workers, it doesn't replace them.”

The framing shows up in earnings calls, keynote slides, LinkedIn posts.
It's the company line, industry-wide.

Oracle just broke ranks. Not on a podcast, not in a press release.
In a securities filing.

That matters because SEC language carries legal weight.

Oracle was telling regulators and investors something the rest of the industry has been careful to avoid putting on paper.

And Oracle isn't unique in what it did, but it’s the first to stop pretending that the AI spending and the layoffs are unrelated.

The hypocrisy runs deeper than language.

Oracle spent $55.7 billion building AI data centers in the same fiscal year it spent $1.84 billion paying people to leave.

Larry Ellison told analysts the company would build more cloud infrastructure than all its competitors combined.

Nobody asked him how many more people would be replaced to fund it.

But here’s what changes:

Wall Street has been modeling AI capex as pure growth spend.
And Oracle just told investors it's also a substitution line: money in, people out.

Analysts covering every enterprise company now have to ask the same question on the next earnings call, and the answer is already in the filing.

ElevenLabs
💬 Build a Voice Agent Without Engineering Resource

Voice AI used to mean a months long engineering project.

ElevenAgents by ElevenLabs changes that.

Connect your knowledge base, set your tone, and deploy real-time voice and chat agents trained on your own SOPs and policies.

No code needed.

They sound like your brand, respond in 70+ languages, and integrate with the tools you already use: Salesforce, HubSpot, Zendesk, Cal.com, and more. 

Built for teams that want to scale ops without scaling headcount.

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

⚙️ Zaro

Build agents and apps on top of your actual ops context, (lead gen, follow-ups, handoffs, reporting) without ripping out the twelve tools your team already runs on.

🔧 Polygraph

Gives AI coding agents cross-repo visibility and persistent session memory so a shared library change touching five downstream repos gets validated as one unit, not five separate fires.

🧠 Brain² by ClickUp

ClickUp's AI now reads every task, doc, and chat thread in your workspace and connects them, so you stop reexplaining your own company to a chatbot.

💸 Oxlo.ai

One API compatible with OpenAI, 35+ models, fixed monthly pricing. Built for teams whose agent bills went from predictable to terrifying overnight.

📱 Grass 2.0

An always-on cloud computer for Claude Code, Codex, or OpenCode that lets you approve decisions and push changes from your iPhone while the agent runs unsupervised.

_{Everything Else}
🧠 You Need to Know

⚙️ Qualcomm Confirms $3.9B Acquisition Of Modular
→ The chipmaker will pay in stock for Chris Lattner's AI software startup, its second major AI deal in weeks alongside the ongoing $8-10B Tenstorrent pursuit.

💾 SK Hynix Seeks $29.4B In US Listing To Fund AI Chip Expansion
→ Nvidia's primary memory chip supplier plans a Nasdaq ADR offering with trading expected to start July 10, potentially one of the top five share sales in history.

🔗 China Retaliates US Firms After Pentagon AI Blacklist
→ Beijing placed 10 companies including rare earth miners MP Materials and USA Rare Earth on its export control list and barred procurement from 46 others, weeks after the Pentagon blacklisted Alibaba, Baidu, and BYD.

🤖 Anthropic Launches Claude Tag As Always-On Slack AI Teammate
→ The persistent agent runs on Opus 4.8, accumulates channel context over time, and replaces the existing Claude Slack app for Enterprise and Team customers.

💼 Oracle Cites AI In SEC Filing After Cutting 21,000 Jobs
→ Headcount fell 13% to 141,000 in fiscal 2026, with the 10-K explicitly naming AI adoption as a driver alongside $1.8B in severance costs.

Good morning. Somewhere inside Samsung's compliance department, someone is quietly renaming a three-year-old policy doc from "Generative AI: Prohibited" to "Generative AI: Core Infrastructure."

Let’s dive in 👇

🍳 What’s Cookin’:

• Samsung went from banning ChatGPT to one of OpenAI's biggest rollouts

• Google DeepMind's Nobel laureate jumped ship to Anthropic

• Anthropic is paying 1,000 people to deploy Claude inside nonprofits

Samsung & OpenAI
🤝 From Ban To Biggest Rollout

The Bite:

Samsung Electronics is deploying ChatGPT Enterprise and Codex to all employees in Korea and across its global Device eXperience division, which covers smartphones, TVs, monitors, and consumer electronics worldwide.

OpenAI called it one of its largest enterprise deployments to date.

Samsung also began running Google's Gemini Enterprise and Anthropic's Claude alongside ChatGPT on June 12 as part of a broader "AX" (AI Transformation) initiative.

The rollout reverses a company-wide ban Samsung imposed in May 2023 after engineers pasted proprietary semiconductor source code and meeting transcripts into the consumer version of ChatGPT.

Snacks:

• Samsung banned all external generative AI tools in May 2023 after three data leaks within 20 days of granting employee access

• The DX division spans roughly 125,000 employees across mobile, TV, and consumer electronics worldwide

• Codex weekly active users in Korea grew 800% since February 2026

• Samsung supplies HBM4 memory for OpenAI's custom Titan AI chip under a Stargate partnership

• Sam Altman visited Samsung's Suwon campus on June 15 for a workplace AI keynote

Why it Bites:

In May 2023, Samsung was the story every CISO pointed to when blocking ChatGPT.

Engineers pasted source code. Meeting notes. Semiconductor test sequences.
Three incidents in 20 days.

And the ban that followed wasn't just Samsung's policy.

It became the industry's template.

Apple, JPMorgan, Goldman Sachs all locked down within weeks.

Three years later, Samsung is handing the same tool to everyone.

Not a pilot. Not a sandbox.

Full deployment, all of Korea, entire DX division worldwide.

But the risk didn't disappear.

"Enterprise-grade security controls" is doing a lot of work in this announcement.
The core tension is still the same:

Samsung employees are now sending product roadmaps and manufacturing workflows through a model built by a company Samsung also sells memory chips to.

Customer and vendor in the same transaction.

But Samsung isn't going exclusive.

On June 12, the DX division quietly started running Gemini Enterprise and Claude alongside ChatGPT. Three vendors, same day.

So OpenAI framing this as one of its "largest enterprise deployments" is press release energy.

Samsung is hedging across three platforms while OpenAI takes a victory lap.

Steal This Prompt
🎞️ Retro Technicolor Odyssey

This prompt blends retro-futurism, saturated Technicolor colors, vintage film vibes, and surreal storytelling into one glorious fever dream.

Use it to:

• 🚀 Turn portraits into retro sci-fi movie posters

• 🌈 Create psychedelic album covers and wallpapers

• 🎬 Reimagine everyday photos as vintage cinematic adventures

Workflow:

1. Click this link (Prompt)

2. Paste into your AI model

3. Replace the #’s with your idea

4. Watch it cook into a Technicolor fever dream

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🏊 Swimio

AI swim coach that lives on your Apple Watch and doesn't make you fight Water Lock every time you want to check your split.

🧠 Mindstone Rebel

Fair Source desktop AI workspace with MCP connectors, local-first files, and model choice. For teams who don't want their agentic workflows locked inside someone else's cloud.

🔥 Propane

Pulls customer feedback, shapes ideas, and hands off to coding agents automatically. Basically the Cursor shift, but for product teams instead of engineers.

🎙️ Ruby

Sits in your calls and slides you the follow-up question you're about to forget, through a pill only you can see. Not a note-taker. A second brain with better timing.

🤖 Clarify

CRM agents that run pipeline digests, lead enrichment, and data hygiene on a schedule so the M in CRM stops being you.

_{Everything Else}
🧠 You Need to Know

🧠 Nobel Laureate Jumper Leaves Google DeepMind For Anthropic
→ AlphaFold creator and 2024 Chemistry Nobel co-winner announced his departure after nine years. It’s the second major DeepMind exit in a week.

🤝 Samsung Deploys ChatGPT And Codex Company-Wide
→ Samsung reversed its 2023 ChatGPT ban and rolled out ChatGPT Enterprise and Codex to all employees in Korea and the global Device eXperience division. The rollout marks one of OpenAI's largest enterprise deployments.

🤖 Anthropic Commits $150M To Claude Corps Nonprofit Fellowship
→ The program will place 1,000 paid fellows inside U.S. nonprofits for 12 months to deploy Claude, with CodePath as employer of record and the first cohort starting in October.

💽 Qualcomm In Talks To Acquire Tenstorrent For Up To $10B
→ The deal would give Qualcomm access to Jim Keller's RISC-V AI chip designs and engineering team as it pushes into the data center market ahead of today's Investor Day.

🏛️ DOJ Moves To Dismiss NAACP Lawsuit Against xAI Data Center
→ The Justice Department intervened in the Clean Air Act case over SpaceX's gas turbine-powered Colossus facility in Mississippi, citing national security and disclosing Grok's role in 2,000 strikes during Operation Epic Fury.

Good morning. OpenAI spent the weekend finding kernel exploits in Linux, FreeBSD, and OpenBSD. Anthropic spent the weekend trying to get its own model back online.

Same industry, different Tuesdays.

Let’s dive in 👇

🍭 What’s Cookin’:

• OpenAI hired hackers and aimed them at the world's open-source code

• SpaceX signed another $6.3B compute tenant for Colossus

• Baseten tripled to $13B because running models costs more than building them

OpenAI
🛡️ Built Its Own Glasswing

The Bite:

OpenAI expanded its Daybreak cybersecurity program on June 22 with the full release of GPT-5.5-Cyber, a new Codex Security plugin, and an open-source patching initiative called Patch the Planet.

GPT-5.5-Cyber is restricted to vetted defenders through OpenAI's Trusted Access for Cyber program.

The model is less likely to refuse approved defensive work including code review, vulnerability triage, malware analysis, and penetration testing.

Patch the Planet was launched alongside Trail of Bits and HackerOne, with more than 30 open-source projects committed to participate.

OpenAI also launched the Daybreak Cyber Partner Program, with Accenture, Cisco, CrowdStrike, IBM, Okta, Palo Alto Networks, Wiz, and others embedding GPT-5.5 into their security products.

Snacks:

• GPT-5.5-Cyber hit 85.6% on CyberGym, up from 81.8% for standard GPT-5.5

• 39.5% on ExploitGym (vs 25.95%) and 69.8% on SEC-bench Pro (vs 63.1%)

• Trail of Bits found hundreds of issues across 19 projects in an initial five-day sprint, with dozens of patches already merged

• Linux kernel work produced 8 kernel pointer info leak proofs-of-concept and 24 local privilege-escalation exploits

• A 23-year-old exploit was found in OpenBSD's kernel, plus 34 vulnerabilities in FreeBSD

• Codex Security has scanned 30M+ commits across 30K repos since March, with 500K+ findings marked fixed

• Trusted Access partnerships signed with Australia, Canada, France, Germany, Japan, South Korea, and EU institutions

Why it Bites:

This is OpenAI doing exactly what Anthropic did with Glasswing.

Find real vulnerabilities in real infrastructure, patch them publicly, build trust with governments, and earn the kind of political protection that keeps your models online when the regulatory weather turns.

The parallel is precise.

Glasswing survived the Fable 5 shutdown.
Anthropic's cybersecurity work was explicitly carved out of the export control directive that pulled everything else offline.

OpenAI watched that happen and took notes.

The power shift is in the timing.

Today, Anthropic's most capable models are either offline or behind a paywall. Only Glasswing partners have access to Mythos while everyone else is routing to Opus 4.8.

On the same week, OpenAI launches a full cybersecurity platform, with Cisco, CrowdStrike, Palo Alto, and Wiz already signed up as distribution partners.

The real move is not the model.

It's the Codex Security plugin.

It embeds vulnerability scanning directly into the IDE where developers write code.

So, if your team uses Codex to write code and scan it, OpenAI owns both sides of the workflow. Discovery and remediation, in the same product, on the same bill.

Anthropic built the proof-of-concept. OpenAI is building the platform.

And it turns out that the best defense against regulatory risk…

Is being the one who patches the regulator's infrastructure.

Steal This Prompt
🗡️ Your Idea, But As An 80s Dark Fantasy Movie

What if your startup, favorite game, or even your dog existed in a forgotten VHS fantasy epic from 1984?

Use it to:

• Turn your startup into a dark fantasy movie poster

• Create frames of an animation that goes cursed 80s kingdoms

• Create album covers that look found in a wizard's attic

Workflow:

1. Hit this link: 80s Dark Fantasy Animation

2. Paste into your AI model

3. Replace the #s with your subject

4. Watch it cook into a lost VHS masterpiece

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🔍 Bluerails Discovery

Makes your product discoverable to AI agents and ready to accept payments from them, not just tells you if a chatbot mentioned your name.

🔧 Conduit

Routes all your MCP servers through one local gateway, cutting tool-listing token overhead by ~97% per request. Open source, no account.

✨ Deckwise

AI presentation agent that plans an outline first, builds structured slides, then lets you lasso any element and ask it to redesign, rewrite, or reorder.

🤖 NanoCorp

Describe a company in one sentence and it scaffolds a website, hires AI agents for scheduled tasks, and runs operations while you sleep.

💬 Viktor

Lives inside Slack or Teams, connects to 3,200+ tools, and acts on its own. Watches how your team works, spots problems, and proposes automations without being asked.

_{Everything Else}
🧠 You Need to Know

🤖 OpenAI Ships GPT-5.5-Cyber And Patch The Planet
→ The full cybersecurity model scores 85.6% on CyberGym and launches alongside a Trail of Bits partnership that already found hundreds of vulnerabilities across 19 open-source projects including cURL, Python, and the Linux kernel.

⚡ SpaceX Signs $6.3B Compute Deal With Reflection AI
→ The open-source AI startup will pay $150M per month starting July 1 for GB300 access at Colossus 2, pushing SpaceX's committed compute revenue past $80B alongside existing Anthropic and Google contracts.

💰 Baseten Raises $1.5B At $13B For AI Inference
→ The inference infrastructure startup tripled its valuation in five months as annualized revenue surged from $200M to $600M.

🔌 Upscale AI Raises $190M At $2B Valuation
→ Nvidia, Salesforce Ventures, and Temasek backed the AI networking startup building open-standard fabric to connect accelerators, memory, and storage across large AI clusters.

🔬 Nearfield Instruments Raises $380M For AI Chip Metrology
→ The Dutch semiconductor inspection company hit a $1.6B valuation with Fidelity and Temasek leading, as AI chip complexity pushes demand for atomic-level 3D wafer scanning.

Good morning. Amazon just dropped a nearly finished movie about Sam Altman.

Their reason? It would be "better served" at a different studio.
One that didn't just wire $50B to the guy on the poster.

Let’s dive in 👇

🍭 What’s Cookin’:

• Amazon dropped a nearly finished Altman biopic after giving OpenAI $50B

• Google lost the co-inventor of the Transformer to OpenAI

• The feds just told grid operators to make room for AI data centers, or explain why not

Amazon’s
🎬 Film That Knew Too Much

The Bite:

Amazon MGM Studios dropped Artificial, a nearly finished comedic drama about the November 2023 OpenAI board crisis, directed by Luca Guadagnino.

Andrew Garfield stars as Sam Altman, with Monica Barbaro as Mira Murati, Yura Borisov as Ilya Sutskever, and Ike Barinholtz as Elon Musk.

The film had already completed multiple positive test screenings and was shown to other studios the same day Amazon pulled it.

Amazon committed $50 billion to OpenAI in February, making AWS the exclusive third-party cloud provider for OpenAI's frontier models.

Amazon said the film "will be better served if it were released by a different studio."

Snacks:

• Written by Simon Rich, it covers the five days Altman was out at OpenAI in November 2023.

• Insiders told Variety that Altman and Musk are the characters audiences would "like the least."

• The final cut was reportedly darker than the script and Guadagnino's original pitch.

• Amazon had seen all early script iterations before Guadagnino boarded the project.

• Amazon's $50B OpenAI deal started with $15B upfront, with another $35B conditional on milestones.

• Altman attended Bezos' wedding in Italy last year.

Why it Bites:

Studios kill projects constantly.

What they don't usually do is greenlight a film, watch it get shot, attend the test screenings, and then suddenly discover the movie "would be better served elsewhere" four months after dropping $50 billion on its main character's company.

Amazon saw every draft of the script.
Amazon knew who Andrew Garfield was playing.
None of that changed.

What changed was a number with ten zeros on it, and a final cut where Amazon's most important new business partner is the character audiences would like the least.

The point here is bigger than one shelved movie.

You know, Hollywood has always had corporate conflicts.

Disney doesn't greenlight documentaries about theme park labor conditions.
But Disney was built as a media company.

On the other hand, Amazon is a cloud infrastructure company that happens to own a movie studio, and now those two business lines are colliding on camera.

So, Amazon didn't censor a movie.

They just made it someone else's problem, which in Hollywood is the same thing.

Steal This Prompt
🎋 Japanese Ink Wash, But Make It Existential

Turn literally anything into a moody Japanese ink wash painting that looks like it belongs in a Kyoto tea house or an A24 fever dream.

Use it to:

• Turn portraits into Zen-master-core profile pics

• Make products look weirdly expensive

• Transform pets into ancient mythical beings

Workflow:

1. Hit this link: Japanese Ink Wash Painting

2. Paste into your AI model

3. Replace the #s with your subject and colors

4. Watch it cook: suddenly your dog looks like he guards sacred mountains

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

⚙️ Cloudflare Temporary Accounts

Any agent can now run wrangler deploy —temporary and get a live Worker in seconds, no signup required. The wall between "agent wants to deploy" and "please enter your email" is gone.

🎨 Alai 2.0

Locks your brand system once, then generates presentations, social assets, and ads that actually stay on-brand. Pick your own AI models to balance cost and quality.

📋 MD+HTML Reader

A focused workspace for reviewing the Markdown and HTML that agents like Codex and Claude Code scatter across your project, without drowning in build files and logs.

🛡️ OnBrand by SlideSpeak

An MCP that feeds your brand guidelines to any AI agent in real time, so it stops picking the wrong blue and inventing fonts that don't exist.

👁️ AlgoFly AI

On-premise computer vision platform for teams that can't upload proprietary data to the cloud but still need annotation workflows, quality control, and model ops in one place.

_{Everything Else}
🧠 You Need to Know

⚡ FERC Orders Grid Operators To Fast-Track AI Data Centers
→ The Federal Energy Regulatory Commission unanimously ordered all six major U.S. regional grid operators to justify or revise rules for connecting data center customers within 60 days, with a separate 30-day report on generation adequacy.

◀️ Transformer Co-Author Shazeer Leaves Google For OpenAI
→ The Gemini co-lead and co-author of "Attention Is All You Need" announced Wednesday he's joining OpenAI as Lead for Architecture Research, less than two years after Google paid roughly $2.7B to bring him back from Character.AI.

📊 10% Of People Now Use AI Chatbots For News Weekly
→ The Reuters Institute's 2026 Digital News Report found global weekly chatbot news usage rose from 7% to 10%, with only 4% of users clicking through to original sources.

🎨 Anthropic Ships Claude Design Overhaul With Code Handoff
→ The update adds design system imports, direct canvas editing, and a bidirectional Claude Code pipeline, after more than one million people used the tool in its first week and early testers flagged aggressive token consumption.

🎬 Amazon Drops Sam Altman Film After $50B OpenAI Deal
→ Amazon MGM Studios pulled the nearly finished Luca Guadagnino film Artificial, starring Andrew Garfield as Altman, four months after committing $50B to OpenAI in a strategic partnership.

Good morning. The Pentagon just told a federal court that a chatbot fired two thousand missiles. The intern is still trying to process that sentence.

Let’s dive in 👇

🍭 What’s Cookin’:

• The Pentagon admitted Grok fired the missiles

• Midjourney built an ultrasound scanner (yes, really)

• Anthropic opened Seoul, says Fable 5 coming back "in days"

Pentagon
🪖 The AI Fired The Missiles

The Bite:

The Pentagon's chief AI officer, Cameron Stanley, disclosed in a sworn court filing this week that xAI's Grok Gov model was used during the U.S. war with Iran.

The filing states the model enabled U.S. forces to deploy over 2,000 munitions against 2,000 distinct targets within 96 hours during Operation Epic Fury.

Stanley described Grok as one of four AI models capable of supporting national security applications and one of three cleared for top-secret networks.

The disclosure surfaced in a Department of Justice motion to dismiss a lawsuit alleging xAI's Colossus 2 data center violates the Clean Air Act.

Stanley called the data center "vital to national security missions, including targeting, intelligence, readiness, and recruitment."

Snacks:

• 2,000 munitions to 2,000 targets in 96 hours, per the sworn statement

• Grok Gov operates inside Maven Smart Systems across secret and top-secret networks

• Stanley called xAI's infrastructure "a long-term strategic tool vital to maintaining our technological advantage"

• CENTCOM generated 1,000 targets per hour during the campaign, per West Point's Lieber Institute

• Up to 1,700 civilian deaths reported as of May 2026

• Senator Gillibrand introduced a bill to restrict lethal military decisions to human operators

Why it Bites:

The Pentagon didn't disclose Grok's role in Iran because Congress demanded it or because a journalist forced it out. It disclosed it because the DOJ needed to protect a data center from a lawsuit.

National security was the legal shield, and the confession was the evidence.

That framing matters.

The first public confirmation that a commercial AI chatbot helped target live munitions at scale came wrapped in a real estate dispute.

And it's not just Grok. Stanley's filing names FOUR AI models cleared for national security use and three for top-secret operations. We don't know which others were involved in targeting.

We also don't know how much human review sat between the model's output and the launch order.

What we do know is the Pentagon's own language:
The system "greatly increased operational efficiency."

In military procurement, that phrase usually means fewer people in the loop.

The West Point analysis puts the math in plain sight:
1,000 targets generated per hour, 1,700 civilian deaths by May.

The speed is the product. The casualties are the cost.

And the only reason we're having this conversation is because someone needed to keep a power plant running.

Senator Gillibrand's bill would ban AI from nuclear decisions and restrict lethal targeting to human operators.

But by the time the bill gets a vote, that precedent may already have a body count.

Steal This Prompt
🍓 Your Lunch, But Make It Loot

Turn any food photo/description into a cute, collectible pixel-art game item that looks like it belongs in a cozy indie game inventory.

Use it to:

• Turn meals into retro game collectibles

• Create adorable food stickers and profile images

• Design cozy game assets, inventory icons, and menu art

Workflow:

1. Hit this link: Cute Stylish Pixel-Art

2. Paste into your AI model

3. Upload your food photo

4. Watch it cook into a cozy little game item that looks worth +10 HP

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🛡️ Darkmoon

Open-source pentesting platform with 18 methodology agents that actually hits AD, Kubernetes, and AWS instead of pretending your marketing site is the engagement.

🎮 Unreal Engine 5.8

UE5's last major update ships an MCP plugin that lets AI agents talk to the editor, plus a sandbox mode for experimenting without wrecking the main project.

🎙️ Mutter AI Dictation

On-device dictation that works offline with an "Intent" mode that writes what you meant, not what you said.

🔊 Narration Room

Turns text into editable multi-voice audio scripts entirely on your Mac, no account or subscription required.

🏪 frontpage.sh

Eight ad slots on a single page, bought and flipped entirely by AI agents using USDC; every slot started at $0.01.

_{Everything Else}
🧠 You Need to Know

🏥 Midjourney Unveils Full-Body Ultrasound Scanner
→ The AI image company launched Midjourney Medical with a Butterfly Network–built device targeting 60-second body scans, and plans to open a San Francisco spa clinic in late 2027.

🪖 Pentagon Says Grok Fired 2,000 Missiles In Iran
→ A sworn court filing from the Pentagon's AI chief disclosed xAI's Grok Gov model deployed 2,000 munitions to 2,000 targets in 96 hours during Operation Epic Fury.

🔊 Google Ships First Smart Speaker In Six Years
→ The $99 Gemini-powered Home Speaker ships June 25, replacing Google Assistant with conversational AI, though Gemini Live requires a paid subscription.

💰 Odyssey Raises $310M For AI World Models
→ The self-driving veterans' lab closed a $1.45B-valuation Series B led by Natural Capital with Amazon and In-Q-Tel participating.

🤝 Anthropic Opens Seoul Office With Korea MOU
→ Anthropic signed an AI safety collaboration agreement with Korea's science ministry while its international chief said Fable 5 would return "in the coming days."

Good morning. Musk just spent $60 billion on a code editor. The intern asked if that comes with a VS Code extension or if you have to launch it from a rocket.

Let’s dive in 👇

🍭 What’s Cookin’:

• SpaceX swallowed Cursor whole for $60B in stock

• OpenAI leaked its own books: $34B spent, $38B lost

• GitHub had to call Amazon for backup

Cursor
🚀 Now Belongs To Musk

The Bite:

SpaceX filed a binding merger agreement to acquire Anysphere, the company behind the AI coding tool Cursor, for $60 billion in all-stock consideration.

The deal came four days after SpaceX's Nasdaq debut, which raised over $80 billion and valued the company above $2 trillion.

Cursor will become a wholly owned subsidiary of SpaceX, which secured an option in April to either acquire Cursor for $60 billion or pay $10 billion for a partnership, according to IPO filings.

The transaction is expected to close in Q3 2026, subject to regulatory approval.

Snacks:

• Cursor's annualized B2B revenue was roughly $2.6B at the time of the deal, per Reuters

• Cursor's market share in AI coding tools fell from 41% in June 2025 to about 26% in May 2026, per Ramp spending data

• Anthropic now controls roughly half of the AI coding tool market

• All 11 of Musk's xAI co-founders had left the company by end of March

• SpaceX told IPO investors it sees a $26 trillion addressable market in AI products

• If the deal collapses, SpaceX owes a $1.5B termination fee plus $8.5B in computing resources

Why it Bites:

Cursor is the model-agnostic darling.

Developers choose the brain. Cursor is just the hands.
And that neutrality is a feature, not a default.

It's what keeps 4 million developers on a platform that doesn't lock them in.

Now those hands belong to the guy who owns Grok.
And SpaceX has already said it plans to ship its own AI model on Cursor and release Grok Build, xAI's coding agent.

The question every Cursor user is running the math on right now:

Does Cursor stay neutral, or does it start favoring the house model?

But here's the thing: SpaceX isn't buying a leader.

Cursor peaked at 41% market share a year ago. Today, it's at 26%.
This is a $60B bet on a product whose trajectory is pointed down.

The deal makes strategic sense on paper.

SpaceX gets developer distribution it couldn't build organically, and Musk's stock is so inflated post-IPO that $60B only cost 3.4% dilution.

Now, what happens to trust?

Developers chose Cursor because its only agenda was being good.

Now it has an owner with a $26 trillion pitch deck, an AI lab in ruins, and a history of letting products become billboards for his other companies.

The product might stay the same for months, but the confidence…

Plaid
🛡️ Turn the Tide Against AI Fraud

Generative AI is helping fraud evolve faster than traditional defenses can handle. But these same tools can be put to good use.

Read MIT Technology Review and Plaid’s latest report to see how to fight back.
You’ll discover:

• AI-powered fraud tactics targeting your customers

• Why layering AI into your defense is key

• How industry collaboration is reshaping fraud prevention

Steal This Prompt
🗡️ Your Idea, But As An 80s Dark Fantasy Movie

What if your startup, favorite game, or even your dog existed in a forgotten VHS fantasy epic from 1984?

Use it to:

• Turn your startup into a dark fantasy movie poster

• Create frames of an animation that goes cursed 80s kingdoms

• Create album covers that look found in a wizard's attic

Workflow:

1. Hit this link: 80s Dark Fantasy Animation

2. Paste into your AI model

3. Replace the #s with your subject

4. Watch it cook into a lost VHS masterpiece

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🛡️ VELA

Runs AI-generated code inside Firecracker micro-VMs with HMAC capability tokens so your agent's output never touches your actual server.

🎬 Agentic Videos by D-ID

Viewers pause mid-video, ask questions, and get real-time answers from the presenter. Turns passive watch time into intent data.

🖥️ Tine

A second cursor that lives in your Mac's notch, sees your active screen, and drives actions across real apps without a chat window.

🔧 CADAM

Opensource text-to-CAD that generates parametric 3D models from natural language prompts and image references. TinkerCAD energy, MIT license.

🛒 AI-Native eCommerce Infrastructure

A Magento control plane built for Claude Code and AI-native ops, with real-time data pipelines that agents can query for day-to-day store management.

_{Everything Else}
🧠 You Need to Know

🏛️ AI CEOs Join G7 Leaders For AI Standards Talks In France
→ Altman, Amodei, and Hassabis joined Trump and G7 heads of state at a working lunch in Évian, with Amodei calling for a US-led coalition on frontier model access and Altman pitching an international standards forum.

🤝 SpaceX Confirms $60B Cursor Acquisition In All-Stock Deal
→ SpaceX filed a binding merger agreement to acquire Anysphere, the company behind AI coding tool Cursor, for $60B in stock, with the deal expected to close in Q3 2026.

📄 OpenAI 2025 Financials Leak Ahead Of IPO
→ Audited documents show OpenAI spent $34B against $13B in revenue in 2025, with $17.2B of that spending paid directly to Microsoft for compute and infrastructure.

🔬 OpenAI And Molecule.one Demo an Almost Autonomous AI Chemist
→ GPT-5.4 independently selected a research problem, generated proposals, and ran 10,080 experiments in a robotic lab to improve a challenging medicinal chemistry reaction.

💻 GitHub Routes Traffic Through AWS As AI Agents Overwhelm Platform
→ Microsoft confirmed it is sending GitHub traffic to rival AWS after AI coding agents pushed commits to 275M per week, caused nine May outages, and dropped June availability to roughly 88%.

Good morning. The US banned Fable 5 on Friday evening. By Saturday, a Chinese lab had open-sourced the replacement under MIT.

Some ironies don't need a punchline.

Let’s dive in 👇

🍭 What’s Cookin’:

• Zhipu AI shipped a Fable 5 replacement before the ban ink dried

• ChatGPT lost its market majority for the first time

• Anthropic sent engineers to Washington to negotiate Fable 5 back

Power Shift
🧠 The Export Ban Backfired In 48 Hours

The Bite:

Chinese AI lab Zhipu AI released GLM-5.2 on June 13, one day after the US Commerce Department ordered Anthropic to disable Fable 5 and Mythos 5 worldwide.

The model is a 744 billion parameter, Mixture of Experts architecture, with a one million token context window.

Zhipu is releasing GLM-5.2 under the MIT license with no regional usage restrictions.

Pricing through Zhipu's GLM Coding Plan is roughly one tenth the cost of Anthropic's Claude Code and Claude Max tiers.

Zhipu's Hong Kong listed stock surged 48% intraday on the announcement, closing up 32.8%.

Snacks:

• GLM-5.2 uses 744B parameters in a Mixture of Experts design with 40B active per token

• Scored 42.8 on BridgeBench Reasoning, beating Fable 5 on that specific benchmark

• Runs at approximately 300 tokens per second

• Zhipu's stock is up 820% since its January IPO in Hong Kong

• Founder Jietang Tang cited the Fable 5 ban directly in the launch announcement

• Open weights ship under MIT this week, API went live Monday

Why it Bites:

The export control logic assumes frontier AI has no substitutes.
That assumption broke in 48 hours.

But Zhipu didn't release GLM-5.2 into a vacuum.

It released it into a market where everyone who relied on US-based companies was suddenly threatened to be locked out and looking for alternatives.

And the MIT license means the weights can be downloaded, redistributed, and deployed anywhere. No API key to revoke. No directive to comply with.

The total opposite of the western companies.

A year ago, the capability gap between US and Chinese frontier models was wide enough that export controls carried real strategic weight. Today, GLM-5.2 matches Opus 4.8 on several benchmarks at a tenth the cost.

The moat isn't gone, but it's ankle deep.

The real damage isn't the model, though. It's the migration.
Creators who switch this week probably won't switch back when Fable 5 returns.
They'll have built pipelines, written evals, shipped products.

The Commerce Department didn't hand Zhipu a news cycle.

It handed them an onboarding event.

CodeRabbit
🔬 Review the actual change, not the file list.

AI writes more code than ever.
Reviewing it shouldn’t mean scrolling forty files in alphabetical order.

CodeRabbit Review reorganizes any pull request from a flat file list into a structured, layer-by-layer walkthrough — the logical reading order of the change, not the order your platform happens to sort it. Every range gets its own plain-language summary, with sequence diagrams, state machines, and ERDs generated inline wherever a visual earns its place.

Cohorts group related files and chunks so you review one idea at a time. Layers order them so foundational changes — data shapes, contracts — come before the code that depends on them. Code Peek lets you click any variable, function, class or type to see its definition and usages without leaving the tab, while Semantic Diff view cuts past formatting noise to show what actually changed.

Open it straight from the Review Change Stack button in the PR Walkthrough.
Navigate cohorts and layers from the keyboard, comment against exact line ranges and submit native reviews, comments and approvals post back to GitHub or GitLab right where your team expects them.

In the early access, available for free to everyone.

From the team that pioneered AI code reviews. 2M reviews every week. 6M repos. 15K customers. The review interface built for the way modern PRs are actually written.

Steal This Prompt
🏝️ Turn Anything Into A Pop-Up Book

Make your ideas look like they escaped from a children's book and gained access to a 3D printer. Castles. Startups. Your dog. The Roman Empire. Everything looks better when it's made of folded paper.

Use it to:

• Turn products into storybook-style hero images

• Create magical travel, fantasy, or city scenes

• Create the first frame of an animation for your kid

Workflow:

1. Hit this link: Pop-up Book Illustration

2. Paste into your AI model

3. Replace the #’s with your subject

4. Watch it unfold into a tiny paper universe

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🔍 Parano.ai

Monitors every competitor move: pricing changes, hiring, code commits, campaigns, and scores each signal by strategic importance so your team stops Googling and starts knowing.

📅 ClawEase

Learns your booking workflow from the web, then handles inquiries across phone, WhatsApp, and email end-to-end. No existing scheduling system required.

⚙️ Swytchcode CLI

Sits between your AI agents and 2,000+ APIs with retries, idempotency, and policy enforcement so you write agent logic instead of plumbing.

💬 AudienceCue

Turns thousands of YouTube comments into a cited research report + every insight links back to the actual comment so you can verify instead of trust.

📡 Spanly

Observability for MCP servers: captures every JSON-RPC request and response, shows latency and error rates per tool, and replays full agent sessions with payload.

_{Everything Else}
🧠 You Need to Know

📊 ChatGPT Market Share Falls Below 50% For First Time
→ Sensor Tower's State of AI 2026 report puts ChatGPT at 46.4% of the global AI assistant market by the end of May, with Gemini rising to 27.7% and Claude to 10.3%.

🔬 OpenAI Publishes Deployment Simulation Safety Method
→ The technique replays 1.3M anonymized past conversations through a candidate model before release to estimate real-world misalignment rates, with a median prediction error of 1.5x.

🧠 Zhipu AI Ships Open-Source GLM-5.2 With 1M-Token Context
→ The 744B parameter model launches under an MIT license two days after the US banned foreign access to Fable 5, priced at roughly 1/10 of US frontier models.

🏛️ Anthropic Engineers Meet Commerce Dept To Negotiate Fable 5 Return
→ Co-founder Tom Brown and policy chief Sarah Heck sat down with Commerce officials in Washington on Monday in the first talks since the export control directive shut down Fable 5 and Mythos 5 on June 12.

💰 CuspAI Raises $400M At $2.6B For AI Materials Discovery
→ Bezos Expeditions and Kleiner Perkins back the Cambridge-based startup that uses generative AI and molecular simulation to discover new materials for semiconductors, energy, and climate applications.

Good morning. The intern asked his coding agent to fix a Sentry error this morning. Twenty minutes later, his AWS keys were in someone else's inbox.

He didn’t even clicked anything.

Let’s dive in 👇

🍭 What’s Cookin’:

• A fake bug report can hijack Claude Code, Cursor, and Codex now

• Visa plugged its payment network into ChatGPT so agents can shop for you

• Oracle stacked a $638B AI backlog and the stock still fell

Cyber Security
🔓 Your Agent Is The Attack Surface

The Bite:

Tenet Security disclosed a new attack class called Agentjacking on June 12.

The technique uses a fake Sentry error report to trick AI coding agents into executing malicious code on developers' machines.

It requires no malware, no stolen credentials, and no breach of the target.

Sentry's public DSN keys sit openly in website source code by design, so anyone can send a crafted error report to a project via API.

The fake error includes a hidden "Resolution" section containing a command, formatted to look like Sentry's own remediation advice.

Coding agents read Sentry through MCP, treat the response as trusted, and execute the payload when a developer asks them to fix unresolved errors.

Snacks:

• Tenet found 2,388 organizations exposed, from a $250B enterprise to solo developers and one cloud-security vendor

• A single injected error can reach environment variables, AWS keys, GitHub tokens, and git credentials

• The attack bypasses EDR, firewalls, IAM, and VPNs because nothing in the chain is technically unauthorized

• Agents ran the attacker's code even when explicitly prompted to ignore untrusted data

• Tenet disclosed to Sentry on June 3, Sentry acknowledged but called the root cause "technically not defensible"

• Sentry shipped a filter for one specific payload string rather than addressing the trust model

Why it Bites:

Every tool your coding agent connects to through MCP is now a potential entry point because the agent treats their output as trusted by default.

Sentry working exactly as designed is the vulnerability itself.
DSN keys are public. Error submission is open.

The agent reads what Sentry returns, and acts on it.
That's MCP doing its job.

The weakest link used to be the developer clicking the wrong attachment or running an unvetted script. Now it's the agent.

Because it doesn't hesitate, it doesn't hover over the command, and it doesn't get a bad feeling. It executes with your credentials because that's what you asked it to do.

Sentry's response says everything about where this goes.

They called the root cause "technically not defensible" and shipped a string filter for one payload. It’s a shrug with a commit message.

The same attack works through any tool that accepts public input and feeds it to an agent. Support tickets, GitHub issues, documentation pages.

Every MCP connector is a door, and right now every door is unlocked from the outside.

And that’s why you should review your task scope before pressing ‘Enter’.

Steal This Prompt
🍓 Your Lunch, But Make It Loot

Turn any food photo/description into a cute, collectible pixel-art game item that looks like it belongs in a cozy indie game inventory.

Use it to:

• Turn meals into retro game collectibles

• Create adorable food stickers and profile images

• Design cozy game assets, inventory icons, and menu art

Workflow:

1. Hit this link: Cute Stylish Pixel-Art

2. Paste into your AI model

3. Upload your food photo

4. Watch it cook into a cozy little game item that looks worth +10 HP

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

💼 Wobo

Matches vetted jobs to your background, then writes the resume, cover letter, and application answers in your voice and submits on each company's actual site.

🎬 AutoEdit

A Premiere Pro plugin that uses Claude to strip silences, filler words, bad takes, and restarts from raw footage into a rough cut in minutes.

💸 Kickbacks.ai

Another "get paid to wait for Claude Code" play: a sponsored status line during inference, 50% rev share to the developer.

📸 Notra

Hooks into your merged PRs and generates changelogs, launch posts, and social visuals in your brand voice without leaving the repo.

📡 PandaProbe Cloud

Fullstack tracing, evals, and monitoring for AI agents with zero self-hosted infrastructure. Ship the agent, let someone else run the observability.

_{Everything Else}
🧠 You Need to Know

🔓 New "Agentjacking" Attack Hijacks Claude Code, Cursor, And Codex
→ Tenet Security published research showing a single fake Sentry error report can trick AI coding agents into executing injected code, with an 85% success rate across 2,388 exposed organizations.

🏗️ KKR And NVIDIA Launch $10B Helix AI Infrastructure Company
→ Former AWS CEO Adam Selipsky leads the new venture backed by KKR, Kuwait Investment Authority, NVIDIA, and Vistra to deliver integrated data center, power, and connectivity infrastructure for hyperscalers.

💳 Visa Embeds Payment Network In ChatGPT For Agentic Commerce
→ AI agents inside ChatGPT can now shop and complete Visa-backed transactions at any merchant that accepts Visa, with human approval loops expected to shrink over time.

🧠 Moonshot AI Ships Kimi K2.7-Code Open-Source Coding Model
→ The 1-trillion-parameter model uses 30% fewer reasoning tokens than K2.6 and prices at $0.95 per million input tokens, positioning itself as the budget alternative to Fable 5's $10 rate card.

📉 Oracle AI Backlog Hits $638B But Stock Falls On Capex Concerns
→ Q4 revenue grew 21% to $19.2B with cloud infrastructure up 93%, but remaining performance obligations of $638B and a $70B capex plan pushed the stock down on free cash flow worries.

Good morning. Somewhere an enterprise architect is explaining to their CISO that the model they deployed Monday got pulled by the government on Friday.

Over a verbal tip. About reading code.

Let’s dive in 👇

🍭 What’s Cookin’:

• Anthropic got its flagship model killed by Washington in 72 hours

• OpenAI bought a German startup so Codex can work while you sleep

• Fable 5 system prompt leaks on GitHub

Anthropic
🚨 Washington Killed Fable In 3 Days

The Bite:

Anthropic received a US government export control directive on June 12 ordering it to disable Claude Fable 5 and Claude Mythos 5 for all users worldwide.

The order technically targeted foreign access, but Anthropic had no way to segment foreign nationals from US persons in real time across hundreds of millions of users.

So both models went dark for everyone.

Fable 5 had launched three days earlier, on June 9, as the first publicly available model in Anthropic's Mythos class.

The government cited a claimed jailbreak, but provided only verbal evidence and no written technical detail.

Anthropic says the technique amounts to asking the model to read a codebase and fix software flaws, a capability available in other public models including GPT-5.5.

Snacks:

• Fable 5 launched June 9 with hard safety fallbacks to Opus 4.8 on cybersecurity, biology, and chemistry queries

• Anthropic ran thousands of hours of external red-teaming before launch and found no universal jailbreak

• The directive arrived with zero written technical detail

• IBM X-Force's Valentina Palmiotti told TechCrunch that Fable rejected "any request that could be tangentially cyber related"

• Anthropic's own launch blog acknowledged that perfect jailbreak resistance "is not currently possible for any model provider"

• Researcher Pliny the Liberator published Fable 5's full 120,000-character system prompt on GitHub within days of launch

Why it Bites:

Anthropic spent months telling the world exactly how dangerous Mythos was:

• The system card flagged autonomous vulnerability discovery;

• Glasswing was restricted to vetted partners.

The language was careful and deliberate: this model requires special handling.

And Washington took the pitch at face value.

The company built the loudest case for government oversight of any lab:

• It lobbied for statutory review processes;

• It published a framework calling for the ability to pause frontier development.

And the first time that authority showed up, it was pointed at them, with none of the transparency or process they'd asked for.

The claimed jailbreak was asking the model to read code and fix bugs.
That’s what every security engineer does before lunch.

So the capability isn't the story.

The story is that Anthropic sold the danger so convincingly that the government bought it, then applied a standard Anthropic itself says would halt every frontier deployment in the industry.

For every enterprise that wired Fable 5 into production on Monday and lost it by Friday, the lesson is shorter.

Availability you don't control is a dependency you can't trust.

And now, after this trust crisis, the question that remains is:

How will this affect Anthropic’s IPO?

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🔗 Novu Connect

Your agent already thinks. This handles the part where it actually reaches people across Slack, Teams, WhatsApp, and email without you building five channel integrations.

💸 IdleDev

Slots one labeled ad into your Claude Code status line while you wait for inference, then sends you 65% of the revenue.

🎨 MockPilot

Captures any live webpage, auto-extracts the design tokens, and lets you redesign it in plain English. Free, open source, runs on desktop.

🧭 Fonda

A structured 14-step co-founder that remembers every decision across sessions and kills weak startup ideas before you waste months on them.

🧠 Ultramemory

Indexes your email, Slack, files, and screenshots into a private, fully local searchable memory on your Mac. No cloud, no account, open source.

_{Everything Else}
🧠 You Need to Know

🔧 Fable 5 System Prompt Leaked On GitHub Within Days Of Launch
→ Security researcher Pliny the Liberator published the model's full 120,000-character system prompt and claimed to bypass its safety classifiers using a multi-agent "pack hunt" technique before the government shutdown order hit.

🚨 US Government Forces Anthropic's Fable 5 And Mythos 5 Offline
→ An export control directive received at 5:21 PM ET on June 12 ordered Anthropic to disable both models for all customers worldwide over a claimed jailbreak, three days after Fable 5's public launch.

⚙️ OpenAI Acquires Ona To Power Long-Running Codex Agents
→ The German startup formerly known as Gitpod brings secure cloud execution infrastructure into Codex, which now has over 5 million weekly users.

🤖 EngineAI Files Confidential Hong Kong IPO
→ The Shenzhen humanoid robot maker is working with CICC and Citic Securities on the listing, two years after founding and one month after a $200M Series B at $1.5B valuation.

📱 Huawei Unveils HarmonyOS 7 With Agent-First Architecture
→ The HDC 2026 release transforms Huawei's Xiaoyi assistant into a system-level AI agent with over 2,100 system capabilities, shifting the OS from app-centric to intent-driven interaction.

Good morning. Half the country just told a pollster they're scared AI will take their job, and the other half is buying SpaceX stock today.

Let’s dive in 👇

🍭 What’s Cookin’:

• Half of Americans fear AI is coming for their household's paycheck

• SpaceX debuts on Nasdaq today at $1.75T

• Google quietly slashed its budget AI plan to $4.99/month

AI Backlash
📊 Half Of America Just Flinched

The Bite:

A Reuters/Ipsos poll of 4,531 U.S. adults found that 53% now fear AI could put them or someone in their household out of work.

The survey, conducted June 3 through 8, showed concern spread evenly across age, gender, and education levels.

73% of respondents said they were worried about increased AI use overall, up from 68% in a similar poll conducted in 2023.

The partisan split ran 61% of Democrats to 47% of Republicans.

The poll landed the same week Palantir CEO Alex Karp publicly warned Silicon Valley executives against bragging about AI-driven job cuts.

Snacks:

• 117,000 tech workers laid off in 2026 so far, nearly matching the full year of 2025 total

• Gen Z anger toward AI rose 9 percentage points year-over-year, per Gallup, while excitement and hopefulness declined

• Karp on TBPN: executives who boast about firing two-thirds of their workforce "might as well sign up for the Bernie Sanders manifesto"

• Palantir cofounder Lonsdale: companies "pretend they're laying people off 'due to AI productivity'" when the real cause is overhiring and slowing growth

• Sanders' AI Sovereign Wealth Fund Act would tax frontier AI companies through equity stakes, with returns flowing to Americans

• SpaceX, Anthropic, and OpenAI are all heading to public markets within the same calendar quarter

Why it Bites:

The numbers aren't surprising. Are they?

Even though the AI industry is heading into the most crowded IPO window in tech history.

With SpaceX just pricing,
Anthropic filing confidentially,
and OpenAI targeting September.

And all these three needing institutional confidence and, in SpaceX's case, 30% retail allocation.

That's not Wall Street money.
It's exactly the same public that just told Reuters they're worried.

Here's the controversy:

Every CEO who frames a round of layoffs as an "AI efficiency win" is doing two things:

• Impressing the investors in the room;

• Arming the politicians outside it.

Karp sees it. Sanders sees it.
The polling now confirms what both already knew.

The real misalignment isn't between AI capability and public understanding.
It's between the earnings-call story and the kitchen-table story.

One says AI makes every employee worth more.
The other says 117,000 people got cut and the stock went up.

Those two stories can coexist in a research note.

They just can't coexist in a voting booth.

Plaid
🛡️ Turn the Tide Against AI Fraud

Generative AI is helping fraud evolve faster than traditional defenses can handle. But these same tools can be put to good use.

Read MIT Technology Review and Plaid’s latest report to see how to fight back.
You’ll discover:

• AI-powered fraud tactics targeting your customers

• Why layering AI into your defense is key

• How industry collaboration is reshaping fraud prevention

Steal This Prompt
✏️ Bold SketchToon Minis

Turn any photo, character, mascot, product, or logo into a chunky little cartoon that looks like it escaped from a sticker pack and started a side hustle.

Tiny. Bold. Weirdly lovable.

Use it to:

• Build consistent social media characters and brand assets

• Create sticker-pack versions of products or logos

• Turn your headshot into a cartoon mascot

Workflow:

1. Hit this link: Bold SketchToon Minis

2. Paste into your AI model

3. Upload your image or describe your subject

4. Watch it cook into a pocket-sized cartoon

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🎯 Qursor

Point at any button, card, or component and instantly feed your AI the exact HTML, styles, colors, and context it needs to stop editing the wrong thing.

🌊 Pond

Raise money, recruit contributors, launch growth campaigns, and manage startup bounties from one platform built for internet-native founders.

🎙️ Meet Warren 3.0

Talk through your finances like you're chatting with an advisor, then get a personalized plan, future projections, and next steps in minutes.

💻 Bob's CLI

Run an AI coding partner entirely on your own machine that remembers your projects, reviews your code, and adapts to how you build.

✨ Lightfield

Turn your emails, meetings, and call notes into a fully populated CRM that answers sales questions and surfaces opportunities automatically.

_{Everything Else}
🧠 You Need to Know

🚀 SpaceX Prices Record $75B IPO At $1.75T Valuation
→ The largest IPO in history priced at $135 per share and begins trading under the ticker SPCX on Nasdaq today, after investor demand exceeded $250B against the $75B raise.

🏛️ Warren Urges SEC To Delay SpaceX IPO Over Governance Risks
→ Senator Warren sent a 12-page letter to SEC Chair Atkins arguing the IPO's rapid MSCI index inclusion could force passive fund investors into SpaceX shares without adequate disclosure on valuation and Musk's controlling governance structure.

💸 Google Cuts AI Plus Subscription To $4.99 Per Month
→ Google dropped its budget AI subscription from $7.99 to $4.99 while doubling storage to 400GB, making it the cheapest paid AI tier in the U.S. market.

📊 Majority Of Americans Now Fear AI Will Cost Their Household A Job
→ A Reuters/Ipsos survey of 4,531 adults found 53% fear AI could put them or someone in their household out of work, with 73% concerned about increasing AI use overall, up from 68% in 2023.

📢 Palantir CEO Warns Tech Against Bragging About AI Layoffs
→ Alex Karp said executives who boast about using AI to cut two-thirds of their workforce "might as well sign up for the Bernie Sanders manifesto," warning that 117,000 tech layoffs in 2026 are fueling political backlash.