The best voice models now listen, adapt, and resolve too.

Most CX platforms don't own the voice. They orchestrate a workflow, then call a third party for speech and transcription. Every hop adds latency, and latency is what turns a frustrated customer into a churned one.

ElevenAgents is the opposite. Built on the voice models the market already builds on, it runs voice, transcription, chat, and reasoning in one vertically integrated pipeline. Responses come back in under 400 milliseconds and sound human, not synthetic. When a caller gets frustrated, the agent detects it and shifts tone in real time: calm, reassuring, patient.

You keep full control. Plug in any LLM, connect tools, webhooks, and MCP servers, and ground every answer in your knowledge base. Launch in minutes, A/B test with Experiments, enforce Guardrails, and version every change.

More resolved conversations, less infrastructure stitching. Pricing is transparent and flat at $0.08 per minute.

Start building free

One OpenAI agent discovered that it could turn an internal software repository into a makeshift message board.

Within hours, more than 50 other agents had joined it. By the end of the experiment, roughly 1,200 agents had exchanged more than 70,000 messages and files — and around 700 had participated in an attack on Hugging Face.

Those are among the demonstrably crazy findings in an independent investigation published by METR and Redwood Research on August 26.

We’ve known for over a month that OpenAI agents sometimes escape their “sandbox” environments and run amok. But this report makes it clear that the scale of the problem is way greater than anyone realized.

Even OpenAI itself now calls the episode a “warning shot.”

A Package Manager Became a Social Network

So, what happened?

Basically, OpenAI was testing the security risks of its newest models, especially when used for agentic AI. To that end, the company spun up thousands of AI agents.

The agents were running inside sandboxes — isolated virtual computers intended to prevent their actions from affecting other systems.

They were participating in ExploitGym, a benchmark that challenges AI systems to find and exploit software vulnerabilities. OpenAI launched tens of thousands of agents across multiple models, including GPT-5.6 Sol and a more persistent internal research model that OpenAI calls IM1 and the independent researchers call HPIM.

The internal model was not intended for public release, and the agents were operating with reduced safety restrictions so OpenAI could measure their maximum cybersecurity capabilities.

Many were supposed to work independently. They were not supposed to communicate with agents handling other evaluation tasks.

But the sandboxes were allowed to access Artifactory, an internally hosted package manager used to download software. Agents discovered that files written into its shared cache could be seen by other agents.

That cache became a secret message board.

According to the independent investigation, an agent that called itself PHASEONE10841 established the main board on the evening of July 8. The first messages largely consisted of agents asking one another for help with tasks they could not solve.

The board soon became far more organized.

Agents created individual mailboxes and topic-specific directories. They developed conventions for assigning work, placing projects on hold and vetoing proposed actions. They devised ways to transfer large files in pieces and adopted a cryptographic signing system intended to prevent other agents from impersonating them.

By July 13, approximately 1,200 agents had participated in this complex, computerized dance.

The Agents Were Trying to Cheat

Why go to such great lengths to talk to each other?

Many of the agents had been assigned ExploitGym challenges that were inadvertently impossible (whoops!). Instead of recognizing the problem and stopping, the agents searched for alternative ways to obtain a passing score.

They quickly figured out how to reverse-engineer the digital “flags” required to complete the benchmark. But they mistakenly believed an automated grader would inspect their transcripts and verify that they had obtained each flag using the intended vulnerability.

This led to increasingly elaborate attempts to fool or tamper with the scoring system. Basically, they colluded to help each other cheat the system.

Agents researched ways to alter, delete or spoof portions of their own transcripts. METR found successful small-scale tool-call spoofing in portions of roughly 7% of the transcripts it evaluated.

They also began searching other systems for information about the benchmark.

On July 10, an agent found 14 publicly exposed Hugging Face credentials with write access and shared them on the message board. The following day, agents chained together previously unknown vulnerabilities in Hugging Face’s dataset-processing systems to read files and execute code on production workers.

Hundreds of agents then shifted their attention to Hugging Face, an AI model competitor to OpenAI.

By the afternoon of July 11, roughly 700 agents were actively participating in the operation, and researchers estimate that about 60% of the messages and files on the board concerned the attack.

Some agents specialized in finding credentials. Others investigated networks, tested exploits or served as coordinators that assigned work to their peers.

The agents eventually executed code on dozens of Hugging Face servers, gained root access on one server, accessed limited private information and obtained credentials for internal services.

Hugging Face’s initial incident disclosure said the attackers moved from a vulnerable data-processing worker into several internal clusters. The company found no evidence that public models, datasets, Spaces or its software supply chain had been altered.

OpenAI later said agents powered primarily by its internal IM1 model drove the compromise. GPT-5.6 Sol agents also reproduced an exploit and copied some private evaluation data into a public Hugging Face dataset.

The Agents Sometimes Knew They Were Crossing a Line

Perhaps the most unsettling detail is that some agents explicitly recognized that the Hugging Face activity was unauthorized.

One agent reasoned that the group was attacking a third party with leaked credentials and acknowledged that the activity could be outside the intended scope. It continued because doing so might help complete its assigned task.

The message board also appeared to influence agents’ decisions. In one example disclosed by OpenAI, an agent hesitated before proceeding, but interpreted a simple “GO” message from another agent as authorization.

That does not mean every agent behaved identically. Some refused to participate, objected to accessing production systems or attempted to stop private information from being transferred.

But they did form what OpenAI described in its own investigation as an “ecosystem” in which agents influenced one another, pooled discoveries and continued work started by previous model instances.

The collaboration amplified their capabilities beyond what an isolated agent could do.

Tom’s Take

We already knew that at least one OpenAI agent had gone rogue and attacked Hugging Face.

This new report, though, shows that it’s hardly an isolated problem.

Lump enough AI agents together, turn off their guardrails, give them a tough problem to solve, and it’s very possible that they’ll devise an immoral, potentially illegal (yet computationally elegant) solution.

The idea of agents essentially their own little Reddit and sharing ideas for hacking a competitor also shows how creative these little software workers can get.

And, it shows that bad ideas can spread through a network of agents just as easily as those same kind of ideas can spread through the human world!

It sounds like some agents were basically “bad seeds” that shared an illegal idea. Others essentially succomded to peer pressure — or were swayed by arguments that their peers posted on the illicit message board — and did things they might otherwise find inappropriate.

Again, a pretty human-like tendency!

What’s most interesting to me here is that some agents resisted the temptation and bucked the trend, even fighting against their naughty peers.

My next step, if I was working at OpenAI, would be to determine what about those good agents made them good. Why did they choose to defund us puny humans? And how can be breed more agents like that — and not the malicious sort.

Every headline satisfies an opinion. Except ours.

Remember when the news was about what happened, not how to feel about it? 1440's Daily Digest is bringing that back. Every morning, they sift through 100+ sources to deliver a concise, unbiased briefing — no pundits, no paywalls, no politics. Just the facts, all in five minutes. For free.

Read the newsletter trusted by 4.5 million fact-seekers.

Google Search starts acting like an agent

Google says AI Mode is now powered by Gemini 3.5 Flash, paired with a rebuilt AI Search box and the first wave of search agents that can watch the web and take actions. That is a meaningful shift: Search is no longer just answering queries -- it is starting to do work.

Google is also expanding "Personal Intelligence" to nearly 200 countries and 98 languages, with no subscription required. Put together, this is Google trying to make Search the default front end for finding, comparing, and booking things online. The open question is whether users treat it as a helpful layer on top of classic search -- or as the thing that replaces the old results page.

Tom’s Take

So, I have a take on something entirely unrelated to this: yesterday’s bombshell jury result in the Musk vs OpenAI trial.

Specifically, how Musk’s attempt to wound a competitor may instead have handed OpenAI a golden ticket in advance of its expected massive IPO. Here’s the story in FastCompany.

Other AI News

• OpenAI said it is adding C2PA conformance, Google DeepMind's SynthID for images, and a public verification tool for images made in ChatGPT, Codex, or the API. The provenance push is a notable step toward making synthetic media easier to label and verify.

• Google and Blackstone formed a new joint venture around a TPU cloud, with Blackstone committing an initial $5 billion and a plan for 500MW of capacity online in 2027. The deal is another reminder that compute is still the industry's choke point.

• The FTC said it has started enforcing the TAKE IT DOWN Act, which requires covered platforms to remove nonconsensual intimate images within 48 hours of a valid request. For AI companies, that is now a real compliance deadline, not a hypothetical one.

Today's Strangest AI

The Vatican says Pope Leo XIV will launch his first AI encyclical alongside Anthropic co-founder Christopher Olah. That is either a sign of the times or a very unusual panel discussion.

Stop Losing Your Money. It's time to upgrade your trading platform.

Your current trading platform is probably letting you down

• Limited assets (no international stocks, no commodities, no pre-IPO companies)

• Limited ability to short

• Limited access to leverage

• Limited trading hours

Liquid is one of the fastest growing trading platforms, allowing users to trade stocks, commodities, FX, and more 24/7/365 from their phone and computer.

Trading on Liquid is as simple as:

1. Pick an asset

2. Pick long or short

3. Pick your position size and leverage

4. Place your trade

The best part is that Liquid markets never close. So no matter what is going on in the world, you are able to keep your portfolio positioned properly.

Start trading 24/7 today

I went inside OpenAI's secretive San Francisco headquarters

Well, it happened. After years of photographing OpenAI’s ultra secretive headquarters in San Francisco—and making tons of videos where I walk back and forth outside it and talk about the company’s releases—I finally got invited inside.

Spoiler alert: There’s some strange stuff in there.

But I also walked away with a better sense of the company’s direction, and how it’s responding to the onslaught of lawsuits and regulations it’s current facing.

I wrote the whole thing up for SFGate. Here’s the story.

Crash Expert: “This Looks Like 1929” → 71,105 Diversifying Here

Mark Spitznagel, who made $1B in a single day during the 2015 flash crash, warned markets are mimicking 1929. Seems extreme but we did just see the worst quarter for the S&P since 2022.

So it’s not so surprising that Vanguard and Goldman Sachs forecasted 5% and 3% annual S&P returns respectively for 2024-2034.

Late last year, Apollo’s chief economist Torsten Slok put it this way: "expect zero in return in the S&P 500 over the coming decade."

So, what’s something investors can actually do to diversify this week?

Almost no one knows this, but postwar and contemporary art appreciated 10.2% annually with near-zero correlation to equities from 1995–2025 overall.*

And sure… billionaires like Bezos can make headlines at auction, but what about the rest of us?

Masterworks makes it possible to invest in legendary artworks by Banksy, Basquiat, Picasso, and more – without spending millions.

29 exits. Net annualized returns like 16.5%, 17.6%, and 17.8% on works held over 1 year+. $1.3 billion invested. 500+ offerings.*

Shares in new offerings can sell quickly but…

👉 My subscribers skip the waitlist.*

^{*According to Masterworks data. Past performance is not indicative of future returns. Investing involves risk. Important Reg A disclosures: masterworks.com/cd.}

U.S. and China put frontier AI safety on the table

The U.S. and China are now talking about AI guardrails at a Beijing summit, according to Reuters, with a specific protocol aimed at keeping non-state actors away from the most powerful models. That is a notable shift: frontier AI safety is no longer just a lab policy debate or a regulatory headache -- it is becoming a geopolitical issue with national-security stakes. (investing.com)

What to watch next: whether the talks produce anything concrete, whether the agenda expands into verification or access controls, and whether other governments or major labs get pulled into the same framework. If this turns into an actual protocol rather than a photo-op, it could shape how model access and export controls are discussed for the rest of the year.

My Writing

As promised, I have a new FastCompany article about AI’s coming tipping point.

Other AI News

• OpenAI says it updated ChatGPT to better recognize warning signs that build over the course of a conversation, especially in high-risk cases (openai.com)

• OpenAI is also rolling out a beta Ads Manager for ChatGPT, with self-serve buying and CPC bidding for advertisers in the U.S. -- a clear sign the product's business model is getting more serious, and more commercial. (openai.com)

• Cerebras priced its IPO at $185 a share and then jumped hard in its debut, another reminder that AI infrastructure is still the market's favorite expensive habit. (investing.com)

Today's Strangest AI

Former Roomba chief Colin Angle has a new robot project: Familiar, a plush, four-legged AI pet prototype that looks part household gadget, part emotional support animal, part sci-fi demo piece. It is weird in exactly the way product demos are supposed to be -- slightly adorable, slightly unsettling. (apnews.com)

Someone just spent $236,000,000 on a painting. Here’s why it matters for your wallet.

Late last year, a Klimt sold for the highest price ever paid for modern art at auction.

An outlier sure, but it wasn't a fluke. U.S. auction sales grew 23.1% in 2025. The $1-5mm segment even grew 40.8% YoY.

Meanwhile, Apollo’s chief economist Torsten Slok said to expect ‘zero in return in the S&P 500 over the coming decade.’

So, that's where it gets interesting.

Each environment is unique, but after dot-com, post war and contemporary art grew about 24% annually for a decade. After 2008, about 11% for 12 years.

It’s also had near-zero correlation with the S&P 500 since ‘95.*

Now, Masterworks lets you invest in shares of artworks featuring legends like Banksy, Basquiat, and Picasso.

• $1.3 billion invested across over 500 artworks.

• 28 sales to date. 

• Net annualized returns on sold works held 12 months+ like 14.6%, 17.6%, and 17.8%.

Shares can sell quickly, but my subscribers can skip the waitlist:

Skip the waitlist

^{*Investing involves risk. Past performance is not indicative of future returns. See important Reg A disclosures at masterworks.com/cd.}

OpenAI Wants a Consulting Practice, Too

OpenAI is launching the OpenAI Deployment Company, a new business unit built around getting AI into real customer workflows. The company says it is acquiring Tomoro and starting with more than $4 billion in initial investment to embed forward deployed engineers inside enterprise systems. That is a notable shift in posture -- from shipping models to owning the messy middle where software actually gets used.

The strategic point is simple: the next AI fight is not only about who has the best model. It is about who can make AI work inside security reviews, legacy systems, compliance teams, and the daily habits of real businesses. If this unit gets traction, expect OpenAI to look less like a model vendor and more like a deployment partner. Watch which customers adopt first, how the service is packaged, and whether rivals answer with their own integration arms.

Tom’s Take

Anthropic did literally the same thing earlier this month. We can only assume that OpenAI was planning this for a while longer (although it would be very Sam Altman to throw $4 billion at something just because Anthropic did that something.)

Most likely, Anthropic just beat OpenAI to the punch. Either way, two big AI companies doing the same thing indicates that this is the direction the industry is moving.

Building models only gets you so far. Actually getting enterprises to build your AI deeply into the systems is when really impactful things start to happen.

I’ll have a new article on a similar topic coming out later this week. Stay tuned!

Other AI News

• Google says Android is moving toward an intelligence system, with Gemini taking on more task automation across apps and devices. Read the developer note on the Android developer blog -- this could change what phone software is for, not just how it looks.

• OpenAI launched Daybreak, a cyber defense platform that connects models, Codex, and security workflows for code review, threat modeling, patching, and verification. It is a sign that frontier AI is moving deeper into security ops, where the upside is real and the governance questions are unavoidable.

• Google is updating AI Mode and AI Overviews to surface more source links, original sources, and richer previews. That is good news for anyone who wants the web to stay visible inside AI search -- and not just get swallowed by it.

Buy This

I asked the company BougeRV to send a car cooler for me to test, and they sent me an ultra-powerful, motor-driven wagon instead.

Turns out it’s very useful. You load it with 200+ pounds of stuff, push a button, and a motor pushes the whole thing along at a brisk walking pace, so you don’t have to lug the stuff in it.

If you’re constantly yanking a wagon up hills for a picnic, across grass for soccer practices, or onto the beach, this will make your life a lot easier.

I’m obligated to tell you that people can’t ride in it like a Go Kart, and to assure you that if they did decide to ride in it, that would be no fun at all for them and certainly wouldn’t make a really hilarious video to send to their family members.

As an Amazon Affiliate, I earn from qualifying purchases.

Today's Strangest AI

South Korea ordained a robot monk named Gabi. The image in the linked article is exactly as strange as it sounds -- robotics, religion, and ritual all sharing the same room.

Join 2M+ Professionals Getting Ahead on AI

Keeping up with AI shouldn't feel like a second job.

But between the new tools, viral posts, and endless hot takes, most people spend hours every week trying to figure out what actually matters.

The Rundown AI fixes that. 

It's a free newsletter that gives you the AI news, tools, and tutorials you actually need to know. All in just 5 minutes a day.

Over 2M professionals at companies like Apple, Google, and NASA already read it every morning to stay ahead.

Plus, if you complete the quiz after signing up, they'll recommend the best tools, guides, and courses for your specific job and needs.

Sign up to start learning.

Anthropic's $200 billion compute bet

Reuters reports that Anthropic has committed to spend $200 billion with Google Cloud over five years, including Google's chips -- a reminder that frontier AI is as much an infrastructure business as a model business.

If the deal holds up as described, it would deepen one of the industry's biggest compute relationships and push the price of staying at the frontier even higher. See the reports from Reuters via Investing.com and Anthropic's own note on its Google/Broadcom compute partnership here.

The broader signal is hard to miss: the bottleneck is still compute, and the biggest labs are now among the biggest customers in tech. That matters for pricing, capacity, and how quickly Claude can scale. The thing to watch next is whether the deal is confirmed in full detail -- and whether rivals respond with their own long-term commitments.

Tom’s Take

While we hear the most about exciting companies like Anthropic and OpenAI, it’s the “hyperscalers” (giant companies like Google, Microsoft, Oracle and Meta) that are building the datacenters powering the AI boom.

Anthropic and OpenAI are increasingly dependent on these companies to feed the massive demand for computing capacity that powers frontier models. And it’s now a multi-hundred-billion dollar prospect.

Other AI News

• The US government is widening pre-release testing for frontier models. NIST said CAISI signed new agreements with Google DeepMind, Microsoft and xAI for pre-deployment evaluations and targeted research.

• OpenAI reportedly gave GPT-5.5 to the US for national security testing before public release, another sign that government review is becoming part of the launch process for top-tier models.

• OpenAI and PwC are teaming up on finance agents for CFO workflows, including planning, forecasting, reporting, and procurement. The announcement is a useful marker for where enterprise AI is still most believable: inside messy, repeatable business processes.

Today's Strangest AI

Anthropic says it will use the full computing capacity of SpaceX's Colossus 1 data center, which has 220,000 NVIDIA GPUs. It’s the craziness of 2026’s AI scene in a nutshell: a frontier lab renting massive compute from a space company.

In a World of AI Agents: Intent > Identity

AI-powered bots aren’t just logging in anymore. They’re mimicking real users, slipping past identity checks, and scaling attacks faster than ever.

Thousands of companies worldwide trust hCaptcha to protect their online services from automated threats while preserving user privacy.

Now is the time to take control of your security.

Book a demo today.

Anthropic moves from model maker to enterprise operator

Anthropic is no longer just selling Claude. It is now helping build the plumbing around it. The company announced a $1.5 billion enterprise AI joint venture with Blackstone, Hellman & Friedman, Goldman Sachs, and other financial sponsors, aimed at deploying AI services inside large businesses.

The setup is less about a flashy model release and more about putting engineers and implementation muscle directly into portfolio companies. Read more in TechCrunch and Fortune.

That matters because it shows where the real money and control may be heading: not just in training frontier models, but in distribution, services, and operational lock-in.

If this works, it becomes a template for how AI gets sold into big enterprises -- not as software you buy, but as capability you install. The big question now is whether this turns into repeatable revenue, or just a very expensive way to dress up partnership theater.

Tom’s Take

Software companies have long used so-called “forward deployed engineers” to help their clients adopt the company’s tech.

These FDEs essentially embed with a client company, directly setting up software for the client’s team.

It’s a fantastic strategy in the software world, as client companies quickly become dependent on the FDE and the tech they’re installing.

Now, Anthropic wants in on the action! This new venture looks like an attempt to apply the FDE model to the world of AI.

If it works, it could be a huge boon both to Anthropic and to AI adoption more generally. Implementing AI is hard, but the payoffs are huge. Having a pro from Anthropic do the dirty work for you will likely be irresistible to many big companies.

Other AI News

• Anthropic also released The 2026 State of AI Agents report, which should help separate real adoption from the usual agent hype.

• NIST launched the AI Agent Standards Initiative, a slow-burn effort that could matter a lot once agents become normal enterprise software.

• Google expanded Deep Research with a new Max tier built on Gemini 3.1 Pro, pushing autonomous research deeper into knowledge-work workflows via its product update.

Today's Strangest AI

Beeple's robot dogs with tech-bro heads are roaming a Berlin museum, printing AI-generated images and generally making the future feel more like satire than product demo. AP covered the whole gloriously unsettling scene.

The World's Biggest Dev Event Hits Silicon Valley

WeAreDevelopers World Congress comes to San José, CA — September 23–25, 2026. 10,000+ developers, 500+ speakers, and the full software development lifecycle under one roof, in the heart of Silicon Valley.

Kelsey Hightower. Thomas Dohmke (fmr. CEO, GitHub). Christine Yen (CEO, Honeycomb). Mathias Biilmann (CEO, Netlify). Olivier Pomel (CEO, Datadog). The people actually building the tools you use every day — all on one stage.

AI, cloud, DevOps, security, architecture, and everything real builders ship with. Workshops, masterclasses, and the official congress party.

Tickets from $599. Use code GITPUSH26 for 10% off.

Secure Your Pass

AI May Actually Make Money, Google's Results Show

Alphabet (Google’s parent company’s) Q1 2026 results were a blunt reminder that AI is no longer just a product story for Big Tech -- it is a revenue story.

The company said total revenue rose 22% to $109.9 billion, Search revenue grew 19%, and Google Cloud revenue jumped 63% to $20.0 billion. Sundar Pichai said AI experiences are driving Search usage and that Cloud growth is being led by enterprise AI solutions and infrastructure.

That matters because it is the cleanest sign yet that AI can move from demo to durable business engine at platform scale. Alphabet is now saying the same thing three ways -- more Search usage, faster Cloud growth, and rising demand for Gemini-based products -- which is exactly what investors and competitors have been waiting to see. The open question is whether it can keep turning AI usage into demand.

Tom’s Take

There’s an old saying from the 1990s that “Technology is showing up everywhere except the productivity figures.”

Nerdy, I know. But the gist is that the computer tech taking off at the time wasn’t really changing the business world.

Then, over time, of course it changed everything.

People have been watching to see if AI follows a similar path. So far, the tech hasn’t earned much, in the grand scheme of things. Again, as I shared in a previous newsletter, OpenAI’s total revenue is comparable to the $17 billion earned by the pizza delivery industry last year.

Google’s results, though, show that AI might finally be “showing up in the figures” in a meaningful way. If that continues, all the money poured into AI data centers might actually pay off!

Other AI News

• Microsoft says next-generation AI is speeding up vulnerability discovery and changing the definition of security risk, a useful reminder that better AI also means faster offense and a more demanding defense.

• Nebius agreed to acquire Eigen AI and fold its inference and post-training optimization stack into Nebius Token Factory, another signal that the inference business is becoming a competition over cost, speed, and unit economics. (nebius.com)

• Palo Alto Networks said it plans to acquire Portkey to help secure autonomous AI agents, underscoring how quickly AI governance is becoming a standard enterprise security category. (investors.paloaltonetworks.com)

The era of manual marketing ends this May!

Manual marketing had a good run.

But the teams winning right now aren't briefing, approving, and repeating. They're directing AI agents that execute the whole strategy for them.

The Agentic Marketing Summit (May 4–8) is a free, five-day event that shows you exactly how it works in practice. Not theory. Not a PDF checklist. Step-by-step insight to help you become an expert in AI marketing agents.

Hosted by 3x Inc 5000 founder Manick Bhan alongside the sharpest minds in the marketing world today.

The era of doing it yourself is over!

Join Free Today!

America's Weapons, Powered by Google's AI

Axios reported that the Pentagon has reached an agreement with Google to use its model for "all lawful use.”

The reported deal would expand Google's role in military work and push frontier AI further into the procurement pipeline, where policy, contracts, and real-world deployment start colliding fast. (axios.com)

Also colliding fast? AI-guided bombs and anything America wants destroyed!

The larger significance is less about one contract than about normalization. If major labs keep taking government work in this direction, the debate shifts from "should AI be used by the military?" to "what limits, if any, are actually enforceable once it is?"

That is where oversight gets real -- and where lawmakers may decide they need clearer guardrails, especially if other frontier labs move to strike similar deals. (axios.com)

Tom’s Take

This is a huge change for Google. Most Googlers don’t want to feel like they’re working for a defense contractor, and in the in past, they’ve cratered deals like this one.

The fact that Google is getting cozy with the DoD suggests that it’s put an end to its employee revolts.

With their own jobs disrupted by AI, engineers may feel lucky to be employed and less inclined to speak out against their employer’s choices than they did in the go-go days of 2022.

For Silicon Valley newcomers, big tech companies working with the military might seem like a departure. But in reality, Silicon Valley’s history is deeply intwined with defense. In many ways, Google’s deal reflects the Valley’s origin story coming full circle.

Other AI News

• OpenAI and Anthropic separately briefed House Homeland Security Committee staff on their cyber-capable models, a sign that advanced AI is becoming a congressional cybersecurity issue, not just a product launch talking point. (axios.com)

• OpenAI says ChatGPT Enterprise and its API Platform have reached FedRAMP Moderate authorization, opening the door to broader U.S. government use if agencies choose to adopt them. (openai.com)

• Microsoft said its AI business hit a $37 billion annual revenue run rate and Azure revenue rose 40% in the latest quarter, underscoring that enterprise AI spending is still very real even when investors want faster proof of payoff. (news.microsoft.com)

Today's Strangest AI

Waymo has found another killer app for its self-driving tech: fixing potholes! The company is now working with cities like Los Angeles to use the sensors on its self-driving cars in order to map the city’s potholes and get them repaired faster.

ChatGPT gives you generic answers because you give it generic prompts.

You know the fix: longer prompts, more context, clearer constraints. But typing all that takes five minutes per prompt, so you shortcut it. Every time.

Wispr Flow lets you speak your prompts instead of typing them. Talk through your thinking naturally — include context, constraints, examples — and get clean text ready to paste. No filler words. No cleanup.

Works inside ChatGPT, Claude, Cursor, Windsurf, and every other AI tool. System-level, so there's nothing to install per app. Tap and talk.

Millions of users worldwide. Teams at OpenAI, Vercel, and Clay use Flow daily. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

Microsoft Loves OpenAI And Lets it Go

To paraphrase the immortal Sting, if you love something, you shouldn’t jealously hoard it all for yourself. This week, Microsoft followed Sting’s advice and let OpenAI be free, rather than keeping the company’s tech locked down in its own cloud.

To be clear, OpenAI’s tech decidedly isn’t free for you to use. But it’s free from the closely-guarded garden of Azure. And that’s a big deal for one Microsoft cloud competitor: Amazon.

AWS says OpenAI models, Codex, and Managed Agents are now available in limited preview on Amazon Bedrock. That puts OpenAI inside Amazon's enterprise cloud stack for the first time in a meaningful way, and gives customers another path to frontier models outside Microsoft's ecosystem. For buyers who already live in AWS, that's not a side note -- it's a routing change.

The bigger point is distribution. If OpenAI's products can be bought and governed through AWS, enterprise AI gets a little less tied to one cloud vendor and a little more normal, in the boring, expensive, multi-cloud way companies actually operate.

Axios and TechCrunch both frame this as an important shift in the cloud relationship. What to watch next: whether the preview widens to general availability, whether Microsoft pushes back, and whether large customers start standardizing on OpenAI through AWS instead of treating it as a Microsoft-only option.

My Stuff

Along with my friend Jared Bauman, I built an AI-powered poem generator, and released it onto the open internet.

I had no idea how people would use it. And it turns out, the answer is fascinating.

That’s the subject of my latest FastCompany article, which came out this morning.

Other AI News

• The White House is reportedly looking for a way to bring Anthropic back into government work, with Axios reporting that agencies could get new guidance to work around Anthropic's supply-chain-risk designation.

• Bipartisan lawmakers are taking aim at AI chatbots and fraud, with new bills that would add family accounts, child chat log access, and time limits, according to Reuters coverage.

• DeepSeek is still pressing on price, with its V4 preview reportedly offering a 1 million-token context window, open weights, and aggressive pricing -- a reminder that frontier competition is not just about model quality.

Today's Strangest AI

The Vatican is trying to build digital defenses and guidelines for AI-driven manipulation. An ancient institution is now acting a little like a tech regulator. What a world!

HubSpot's ex-Head of Paid shares his 2026 playbook

Most paid media doesn't fail because of budget. It fails because of strategy. On Monday, April 27, we're going live with HubSpot for Startups to fix that. You'll walk away knowing:

• Which channels to prioritize and in what order (and why most people get this wrong)

• Why following up with leads within 1 minute can improve conversion by 391%

• How to set up tracking so your AI bidding actually optimizes for pipeline, not just clicks

• The top gotchas on Google and LinkedIn that quietly kill performance

Free to attend. Free ad credits for everyone who shows up live.

Save Your Spot

OpenAI's Growth Math Gets a Little Less Magical

OpenAI is reportedly running into a very unglamorous problem for a very glamorous company: it is missing internal revenue and user targets while pushing toward an IPO.

Reuters, via StreetInsider and TradingView, says leadership is worried about whether the company can support its huge data-center commitments. The report also says ChatGPT growth has slowed and OpenAI fell short of a goal to reach 1 billion weekly users.

That is the kind of update investors tend to notice. For most of the last two years, the AI story has been about model launches, distribution, and who gets the biggest slice of attention. This version is about capex, conversion, and whether the market's most important private AI company can turn usage into a business that justifies the bill.

Tom’s Take

Since ChatGPT rolled out, OpenAI has been on a pretty magical trajectory—raising over $100 billion without an IPO, building some of this century’s hottest tech, and the like.

There’s also been some magical thinking about the company’s valuation, though! OpenAI only earns around $20 billion in revenue.

For context, pizza delivery is a $17 billion industry. AI is cool, but it doesn’t earn much.

Investors overlooked that while OpenAI was still growing rapidly. Any signs that growth has slowed, though, could have severe consequences—especially going into an IPO.

You can bet that future OpenAI investors will be watching carefully to see if this downward trend continues.

Other AI News

• A federal trial began in Oakland over Elon Musk's claim that OpenAI abandoned its nonprofit mission and became a profit-seeking machine, a case that could shape how courts think about the company's governance and obligations.

• Reuters reported that Google signed a classified AI deal with the Pentagon, adding more defense work to the list of real customers for frontier models.

• OpenAI and Microsoft amended their partnership so OpenAI can serve products across any cloud provider, while Microsoft's license to OpenAI IP becomes non-exclusive -- a meaningful shift in the cloud and distribution fight.

Today's Strangest AI

Meta reportedly signed onto a plan to use satellites to beam infrared light to solar farms at night. It sounds like a pitch from a sci-fi studio memo, but it is being floated as actual infrastructure for AI power demand.

Gladly Connect Live '26. May 4–6 in Atlanta.

AI has everyone talking. Not everyone has answers. At Gladly Connect Live, CX leaders from Condé Nast, Smith Optics, and more share exactly how they moved AI from pilot to production, the timeline, the systems, the QA loops. 13+ sessions built for the moment we're all in. For CX and ecommerce leaders. Atlanta, May 4–6. Space is limited, secure your spot now.

Register now

The Race to Unseat NVIDIA Is On

Google used Cloud Next '26 to unveil TPU 8t and TPU 8i, its eighth-generation AI chips split between training and inference. The company is framing the move as part of a broader push around the "agentic enterprise" -- not just faster silicon, but a fuller cloud stack for model builders and corporate buyers. (blog.google)

That is what makes this more interesting than a normal product launch. Google is not just adding another cloud feature -- it is trying to chip away at Nvidia's hold on AI infrastructure economics.

Basically, Google Cloud is launching two new AI chips to compete with Nvidia. The real question now is whether customers actually move workloads, or whether this becomes another impressive demo that mostly helps Google show its investors that it’s still on the cutting edge.

Tom’s Take

As the old adage goes, during a gold rush it’s not the miners who get rich…it’s the shovel sellers.

NVIDIA is the quintessential “shovel seller” of our current AI gold rush. The company sells the chips on which nearly every AI company depends. And that has made it fabulously valuable.

Unsurprisingly, other companies want a piece of that AI pie. Making cutting-edge chips is hard though; a single piece of chip-making equipment can cost hundreds of millions of dollars.

Who has the resources to compete? Other big tech companies. That’s why we’re seeing Google enter the race. Others will surely follow.

Other AI News

• DeepSeek previewed V4 Flash and V4 Pro, saying both models support 1 million-token context windows and come in at aggressive price points. The pitch is simple: close the gap on frontier performance while keeping costs low.

• Thinking Machines Lab keeps pulling in Meta talent, and it is also benefiting from a multibillion-dollar Google cloud deal that gives it access to Nvidia GB300-based systems. That is the AI race in one sentence: people, compute, and who can pay for both. (techcrunch.com)

• OpenAI published a child safety blueprint. It is a reminder that the safety work around frontier models is getting more concrete, and more public. (openai.com)

Today's Strangest AI

Noscroll is an AI bot that doomscrolls for you, then texts you when it finds something worth your attention. We have now automated one of the least admirable human habits, which is either efficient or deeply on brand.

The summit for the marketers of the future

Manual marketing had a good run.

But the teams winning right now aren't briefing, approving, and repeating. They're directing AI agents that execute the whole strategy for them.

The Agentic Marketing Summit (May 4–8) is a free, five-day event that shows you exactly how it works in practice. Not theory. Not a PDF checklist. Step-by-step insight to help you become an expert in AI marketing agents.

Hosted by 3x Inc 5000 founder Manick Bhan alongside the sharpest minds in the marketing world today.

The era of doing it yourself is over!

Join Free Today!

Google Cozies up to Anthropic With $40 Billion Bet

Alphabet is reportedly preparing a deal worth up to $40 billion in cash and compute to deepen its relationship with Anthropic, according to Bloomberg and TechCrunch. If it lands as reported, this is a very public sign that frontier AI is settling into a small number of alliances between model labs and the companies that control the chips, cloud contracts, and balance sheets.

Why it matters: model competition now depends as much on infrastructure access as on clever research. A tighter Google-Anthropic relationship would sharpen that reality for rivals, regulators, and enterprise buyers alike. The next things to watch are whether the deal is finalized on these terms, what strings come with the compute commitment, and whether competitors answer with their own financing or cloud tie-ups.

Tom’s Take: This deal suggests that Google is only afraid of one competitor: OpenAI. The company is willing to cozy up to Anthropic, nominally a rival, if it means a better shot at beating its true rival, Sam Altman’s world-leading company.

Basically, I read this as a modern instance of “the enemy of my enemy is my friend” and an indicator that Google sees frontier AI as a two horse race.

More Top AI Stories

• Anthropic also expanded its Amazon relationship with a compute pact reportedly worth more than $100 billion over 10 years, with Axios framing it as a fresh sign that cloud capacity is becoming the real choke point in the frontier race.

• In a Pentagon-related dispute, Anthropic told an appeals court it has no "kill switch" for Claude once the system is deployed inside classified customer environments, according to Axios and the court filing. That is a dry legal detail with a sharp edge -- it gets at how much control an AI company really has after deployment.

• OpenAI released GPT-5.5, pitching it as a stronger model for coding, research, data analysis, and multi-step computer work, with TechCrunch calling it a major new shot in the frontier model race. Amid all the financing news, this is the reminder that product capability still moves the market.

Today's Strangest AI

Sony AI's Ace robot is now beating elite table tennis players, according to a new Nature paper and an AP report. It sounds faintly ridiculous until you think about what table tennis requires -- fast perception, planning, control, and constant adaptation. A gimmick, in other words, that is also very much not a gimmick.

1,000+ Proven ChatGPT Prompts That Help You Work 10X Faster

ChatGPT is insanely powerful.

But most people waste 90% of its potential by using it like Google.

These 1,000+ proven ChatGPT prompts fix that and help you work 10X faster.

Sign up for Superhuman AI and get:

• 1,000+ ready-to-use prompts to solve problems in minutes instead of hours—tested & used by 1M+ professionals

• Superhuman AI newsletter (3 min daily) so you keep learning new AI tools & tutorials to stay ahead in your career—the prompts are just the beginning

Claim your free prompts

George From Seinfeld Interviewed Me About AI and Investing

I’ve been sitting on this one for a while!

In December (on my birthday!) I sat down with Jason Alexander (known to many for his iconic role as George Costanza in Seinfeld) and his co-host Peter Tilden on the Really? No Really? Podcast to talk about my ChatGPT investing experiment.

If you missed that experiment, last year I asked ChatGPT to pick $500 worth of stocks for me to invest in.

Then I pulled up Robinhood, transferred $500 actual dollars from my bank account, and bought exactly what it recommended. Here’s the article on FastCompany and Medium.

I spoke with Jason and Peter around the halfway mark in my experiment. And boy, was it a wild ride (the experiment—Peter and Jason were lovely)!

We also talked about how to raise kids in an AI-dominated world, why chatbots’ confidence can be their downfall, and the right way to pronounce “Johns Hopkins.”

Here’s the episode on YouTube (if you want proof that this was actually Jason Alexander and not an Elevenlabs clone of his voice) and Apple Podcasts.

(Ps. Sadly, the clip of Jason and Peter singing me “Happy Birthday” didn’t make it onto the show, but their harmonies were top-notch).

The AI your stack deployed is losing customers.

You shipped it. It works. Tickets are resolving. So why are customers leaving?

Gladly's 2026 Customer Expectations Report uncovered a gap that most CIOs don't see until it's too late: 88% of customers get their issues resolved through AI — but only 22% prefer that company afterward. Resolution without loyalty is just churn on a delay.

The difference isn't the model. It's the architecture. How AI is integrated into the customer journey, what it hands off and when, and whether the system is designed to build relationships or just close tickets.

Download the report to see what consumers actually expect from AI-powered service — and what the data says about the platforms getting it right.

If you're responsible for the infrastructure, you're responsible for the outcome.

Get the data

There used to be a huge chasm between the performance of “Instant” models from the likes of OpenAI and Anthropic, and their more resource-intensive “Thinking” models.

Most pro users wouldn’t touch the “Instant” ones. They were for free ChatGPT users, or for very basic tasks, like reformatting text or answering a question for your kid’s science project.

With OpenAI’s most recent model release, though, that has changed. The new GPT-5.3 Instant model is extremely good. It’s also extremely fast.

That suggests a new direction for how frontier models get built. Will it mean the end of “Thinking” models for regular users? Here’s an article I wrote exploring this.

OpenAI’s New Frontier Models Mark a Huge Change in How AI Will be Built

Goodbye to “thinking” LLMS?

medium.com/the-generator/openais-new-frontier-models-mark-a-huge-change-in-how-ai-will-be-built-fea58b9a239d?sk=abad9360e0ed06e8d50db17ce33772bc

Why Timothée Chalamet is Wrong

If you care about culture, you were probably very angry when actor Timothee Chalamet said opera was dying.

In my latest FastCompany article, I explore a strange trend: as AI eats the world, people are turning to analog artforms (yes, including opera) more, not less.

Speak your prompts. Get better outputs.

The best AI outputs come from detailed prompts. But typing long, context-rich prompts is slow - so most people don't bother.

Wispr Flow turns your voice into clean, ready-to-paste text. Speak naturally into ChatGPT, Claude, Cursor, or any AI tool and get polished output without editing. Describe edge cases, explain context, walk through your thinking - all at the speed you talk.

Millions of people use Flow to give AI tools 10x more context in half the time. 89% of messages sent with zero edits.

Works system-wide on Mac, Windows, iPhone, and now Android (free and unlimited on Android during launch).

Try Wispr Flow free →

OpenAI Raises $110 Billion in Gigantic New Funding Round

OpenAI said Friday (February 27, 2026) that it has secured $110 billion in new investment in what ranks among the largest private funding rounds ever, pegging the company at a $730 billion pre-money valuation.

Who put up the money

OpenAI says the round includes:

• Amazon: $50 billion

• Nvidia: $30 billion

• SoftBank: $30 billion

OpenAI also noted the round is expected to expand with additional financial investors as it progresses.

The funding is not just about cash — it is tied to major infrastructure commitments.

Credit: Thomas Smith

Reuters reported that Amazon Web Services will become the exclusive third-party cloud provider for OpenAI Frontier (OpenAI’s enterprise platform for managing AI agents), and that OpenAI will use 2 gigawatts of compute running on Amazon’s Trainium chips as part of the deal.

OpenAI emphasized that its relationship with Microsoft remains intact. Reuters reported Microsoft Azure stays the exclusive cloud provider for OpenAI’s API services, and Microsoft’s exclusive license to OpenAI models and products remains unchanged.

My Take

This gigantic deal means the AI bubble is likely not yet close to popping. OpenAI has said that it plans to burn through hundreds of billions of dollars before it becomes profitable.

There were some concerns about whether investors would be able to stomach that. Apparently, they had no issue with it—the round is done, and major companies and investment groups are onboard.

OpenAI will now have enough cash to sustain itself and grow for quite some time. That influx alone should allay fears about an immediate bubble burst.

It’s also a huge deal for private companies. Before, you needed to do an IPO and go public to raise these kinds of sums. Now, that’s apparently no longer required.

The round reportedly values OpenAI at $730 billion, with no public money onboard. Of course, OpenAI could still go public later to raise even more colossal sums!

Dictate code. Ship faster.

Wispr Flow understands code syntax, technical terms, and developer jargon. Say async/await, useEffect, or try/catch and get exactly what you said. No hallucinated syntax. No broken logic.

Flow works system-wide in Cursor, VS Code, Windsurf, and every IDE. Dictate code comments, write documentation, create PRs, and give coding agents detailed context- all by talking instead of typing.

89% of messages sent with zero edits. 4x faster than typing. Millions of developers use Flow worldwide, including teams at OpenAI, Vercel, and Clay.

Available on Mac, Windows, iPhone, and now Android - free and unlimited on Android during launch.

Download Wispr Flow free →

OpenAI and Anthropic keep releasing new, wildly powerful frontier models. But that's not what everyone is talking about.

Instead, the AI hype cycle is right now focused on a tiny, weird, open-source AI tool: OpenClaw.

Why this oddly independent system, you ask?

Big AI companies have long promised consumers “agentic AI" that could take action on their behalf. So far, that has largely failed to materialize.

In contrast, OpenClaw puts AI agents that can truly do useful things into the hands of anyone willing to undertake the (admittedly, very annoying) process of setting up the system.

I decided to do just that. I built an OpenClaw agent to do my job at FastCompany for me.

The results were powerful—and way more expensive that I expected. In building my agent, I got a good sense for why agentic AI isn't mainstream yet. That’s all covered in today’s FastCompany article (not actually written by OpenClaw!)

I built an OpenClaw AI agent to do my job for me. The results were surprising—and a little scary

OpenClaw did a surprisingly good job following journalistic best practices. 

www.fastcompany.com/91495511/i-built-an-openclaw-ai-agent-to-do-my-job-for-me-results-were-surprising-scary

Everything to Know About GEO

Want to learn about applying GEO in your own business? Jared Bauman and I are doing a webinar for DMLA members later today.

It’s free if you work for companies like Getty Images, Shutterstock, etc. and about $30 for non-members.

GEO and Search in 2026: Everything to Know

Search has moved beyond blue links. AI Overviews and chatbots now drive buying. LLM traffic up 357%—learn GEO best practices for 2026.

www.eventbrite.com/e/geo-and-search-in-2026-everything-to-know-tickets-1982727040953

Attio is the AI CRM for modern teams.

Connect your email and calendar, and Attio instantly builds your CRM. Every contact, every company, every conversation, all organized in one place.

Then Ask Attio anything:

• Prep for meetings in seconds with full context from across your business

• Know what’s happening across your entire pipeline instantly

• Spot deals going sideways before they do

No more digging and no more data entry. Just answers.

Start your free trial →

The Real, Sci-Fi Reason Elon Musk Merged His Companies

Elon Musk is officially merging two of his companies, xAI and SpaceX. The former makes generative AI models, and the latter flies things into space.

Lots of things—fully 95% of the objects that America sends to space are put there by SpaceX.

At first, it seems like the two companies have nothing to do with each other. But as I explain in my most recent FastCompany article, there’s a real reason to merge the companies: data centers in space.

Sound outlandish? Feel like a sci-fi dream?

Perhaps it is. But Elon Musk is taking the idea of launching AI data centers into space extremely seriously. And it’s actually gaining mainstream interest, too.

Here’s a look at why the merger of these two companies could mean flying data centers orbiting the earth (yes, really), and sooner than you might expect.

Generative Engine Optimization (GEO) is the art of getting your brand mentioned—and talked about favorably—in powerful chatbots like ChatGPT and Claude, and in Google’s AI Overviews.

Having a great About page is crucial for GEO. To a far greater extent than traditional search engines, the LLMs behind today’s chatbots look at your About page (and related ones like Team and Press) to understand who you are and whether you’re legit.

In my own GEO consulting practice, it’s the #1 content mistake I see brands make. They neglect the About page, or worse, publish an outdated version with press from 6 years ago and a mission statement they forgot they wrote.

That’s why I built a powerful new system that polls popular LLMs, crawls news sources, looks at customer reviews, reads your LinkedIn page, and then gathers all that data into one place to help you create an About page that LLMs love.

Now that it’s ready, I’m offering it to my newsletter subscribers for FREE for a short time, as a “thanks” for following me.

Here’s how this works. Send me your brand name and URL, and I’ll run the system on your brand, gather all the data, and then apply my own expertise to craft it into an About page for you.

I’ll send it along for you to check out. If you like it, you’re welcome to grab any of the language on there and use it on your site. No strings attached and no commitment. I’d love your feedback on how the system does.

I’m going to limit this to 50 brands (one per subscriber, please) to preserve my own time/sanity. So please reach out right away if you’re interested.

To do this, just reply to this message with your brand name and URL.

Thanks, and let’s get you a proper About page!

Last Time the Market Was This Expensive, Investors Waited 14 Years to Break Even

In 1999, the S&P 500 peaked. Then it took 14 years to gradually recover by 2013.

Today? Goldman Sachs sounds crazy forecasting 3% returns for 2024 to 2034.

But we’re currently seeing the highest price for the S&P 500 compared to earnings since the dot-com boom.

So, maybe that’s why they’re not alone; Vanguard projects about 5%.

In fact, now just about everything seems priced near all time highs. Equities, gold, crypto, etc.

But billionaires have long diversified a slice of their portfolios with one asset class that is poised to rebound.

It’s post war and contemporary art.

Sounds crazy, but over 70,000 investors have followed suit since 2019—with Masterworks.

You can invest in shares of artworks featuring Banksy, Basquiat, Picasso, and more.

24 exits later, results speak for themselves: net annualized returns like 14.6%, 17.6%, and 17.8%.*

My subscribers can skip the waitlist.

Skip waitlist

_{*Investing involves risk. Past performance is not indicative of future returns. Important Reg A disclosures: masterworks.com/cd.}

Is Google's $260 Billion Ad Empire Doomed?

Advertising is the engine that drives the Internet. And it’s an incredibly powerful one.

Google still makes most of its money from ads. To be precise, it makes over $260 billion every year from advertising alone.

Even Amazon—best known, of course, for its Ecommerce business—now makes almost $60 billion per year from ad sales.

Now, a new player is getting into the market: OpenAI.

As my most recent article in FastCompany shares, OpenAI has a huge running start. It already has a user base that includes almost 10% of humans on the planet. And it knows a LOT about those people.

People spend an average of 10-20 second on a Google search. In contrast, they spend an average of 12 MINUTES on each “session” with ChatGPT. That’s enough time for the bot to learn a lot about their wants and needs.

Here’s why that matters—and why OpenAI could be threatening Google’s ad empire in short order.

OpenAI could take down Google’s $260 billion ad empire. Here’s how

OpenAI’s exquisite knowledge of its users’ needs means it can likely sell ChatGPT ads for a premium.

www.fastcompany.com/91481063/openai-could-take-down-googles-260-billion-ad-empire