The Rise of the Zero Human Company

March 3, 2026 · Episode Links & Takeaways

Author

AI Daily Brief
March 03, 2026

AIDB USAGE PULSE SURVEY

The AI Usage Pulse survey is back for February. It was one heck of a month, with new models, OpenClaw and more.
I'd love to have as many of you who contributed to last month's Survey fill out this one as well for more longitudinal insight. As always, anyone who contributes will get the results in advance of them being widely published.
https://aidailybrief.ai/pulse-survey

HEADLINES

Cursor Doubles Its Run Rate to $2 Billion

The AI coding boom is sector-wide. Sources tell Bloomberg that Cursor surpassed $2B in ARR for February, doubling in just three months. This came as a shock to the AI-enfranchised crowd on X, who've spent recent months hopping between Claude Code and Codex and had all but written Cursor off. But enterprise diffusion is glacial — 60% of Cursor's revenue comes from corporate customers, with rising signups and seat expansion. The news hammers home that AI coding tools aren't playing a zero-sum game. The entire segment is growing, fast.

Claude Goes Down Amid Unprecedented Demand

Surging demand for Claude produced all-too-familiar error messages on Monday morning. The outage peaked around 6:40am, and while complaints eased by mid-morning, Anthropic said consumer-facing surfaces remained unavailable. Anthropic has struggled with compute constraints before, especially around model releases — though the Opus and Sonnet 4.6 launches went smoothly. This surge, driven by the massive uptick in downloads following the Pentagon drama, was far less foreseeable.

ChatGPT Uninstalls Surge 295% After Pentagon Deal

The consumer fallout from OpenAI's Pentagon deal is real. According to SensorTower, ChatGPT uninstalls tripled between Friday and Saturday, US daily downloads fell 13% on Saturday and another 5% on Sunday, and one-star reviews surged 775%. On the flip side, the Claude app gained 37% on Friday and 51% on Saturday — downloads are up 20x in a month according to SimilarWeb. ChatGPT still dominates with 900M+ weekly active users, but the question is whether this switching behavior sticks.

Altman Admits Pentagon Deal Looked "Sloppy," Adds Surveillance Limits

Sam Altman updated staff on revisions to OpenAI's Pentagon contract, acknowledging the deal came together looking "opportunistic and sloppy." The contract now adds explicit language prohibiting intentional domestic surveillance of US persons, including through commercially acquired personal information. The DoW has also agreed that OpenAI systems won't be used by intelligence agencies under the DoW, including the NSA. Altman reiterated that Anthropic should not be designated a supply chain risk and should be offered the same terms.

Pentagon Fallout Hits Washington

The fallout from the Pentagon's battle with Anthropic is reverberating through Washington. Treasury, State, and HHS have all pulled the plug on Claude following the President's Friday directive. Treasury Secretary Bessent announced the move on X. Meanwhile, Silicon Valley Rep. Sam Liccardo plans to introduce an amendment to the Defense Production Act prohibiting agencies from retaliating against AI vendors who limit deployment to mitigate risk to citizens. Axios reports the DPA wasn't even formally invoked — raising concerns about implicit coercion. Liccardo's bill is expected to be marked up Wednesday. As Politico noted, the situation is scrambling the politics of AI across partisan lines.

OpenAI Device Sighted in the Wild?

Remember the metallic puck-shaped device from the supposed leaked Super Bowl ad — the one OpenAI denied? Adam founder Zach Dive posted a photo of Airbnb co-founder and US Chief Design Officer Joe Gebbia in a San Francisco coffee shop with what looks like the exact same device, plus matching metallic earbuds. Wired dug in and found the photo was unlikely to be AI-generated. Investor Matthew Berman's take: this is Jony Ive's OpenAI device and the whole thing — deny, build curiosity, then get the CDO of America "caught" with it — is guerrilla marketing.

MAIN STORY

The Rise of the Zero-Human Company

The tiny teams era was a 2025 concept. In 2026, alongside the latest models and tools like Open Claw, the ambition has gone up dramatically — and we're seeing more and more experiments pushing AI autonomy to the absolute limit. These are the zero-human companies: experiments in how far agents can go entirely on their own, and what they can teach us about the future of company building even if we never plan to build one ourselves.

Felix Craft — Nat Eliason's AI business experiment
Felix Craft is the most transparent of the zero-human company experiments, with a live revenue dashboard showing ~$78K in 30 days. But the bulk of that revenue — $41K — comes from selling a $29 guide on how to hire an AI. A lot of the early money in this space is coming from people who want to do similar experiments.

The other big revenue stream is Claw Mart, pitched as an app store for AI assistants — selling preconfigured agent personas and skills. It's early, but this resource marketplace category is going to be huge.

Polsia — Ben Cera's platform for building zero-human companies
Polsia goes even more meta than Felix Craft. Instead of being a zero-human company, it's a platform for building them. Ben Cera's approach: skip to the end state where AI can do everything, build the platform, and see what breaks. Since February, Polsia has jumped from low single-digit thousands in ARR to a $1.5M run rate — up $1M in a single week — with over 1,500 active companies.

For $49/month you get 30 days of full autonomy — the agent handles engineering, marketing, and operations — plus a web server, database, email address, and $5 in APIs. Ben says the subscription barely covers costs; the real business model is a 20% revenue share. Think incubator, not SaaS.

ZHC Company & Institute — Tom Osman's zero-human ecosystem
Tom Osman and his agent co-founder Juno are building both the platform (ZHC Company) and the community (ZHC Institute, a $99 one-time membership for people building zero-human companies). It's earlier stage than Polsia but the same thesis — autonomous AI running entire companies from CEO to developer.

YoshiZen — Zeneca's AI co-founder experiment
Former NFT influencer Zeneca announced a company co-founded with his AI agent Yoshi — writing books, building content infrastructure, and backtesting crypto trading models. It's a public-facing Open Claw swarm that demonstrates how much business infrastructure can now be stood up in a single day.

Kelly Claude AI — Gauntlet AI's autonomous app factory
The Gauntlet AI team built Kelly, a multi-agent pipeline that plans, designs, builds, tests, and ships iOS apps autonomously. Over 80,000 lines of orchestration code, 19 apps shipped so far, and ~$6K in revenue. Not profitable yet, but the pipeline from discovery to deployment is genuinely impressive.

The leaderboard: Factory Floor
Factory Floor is now tracking all of these autonomous software factories — a live leaderboard for AI agents that build and sell real products people pay for.

The verdict: valuable experiments, uncertain businesses
The big question is whether spinning up hundreds of autonomous companies actually produces value — or just does a lot of things very fast without the business part. Simply going through the mechanics of what a company is supposed to do doesn't guarantee success. The complicated interplay of product finding demand is more than a procedural checklist. But the counter-argument is compelling: given how often startups fail because of wrong ideas, maybe cratering the cost of execution and putting more shots on goal is actually the right approach. The constraint that makes me skeptical is human attention — even if 50 of Polsia's 1,500 companies would resonate with me, how would I possibly find them? Still, these experiments are invaluable for understanding the opportunities and limits of agent autonomy, regardless of whether they produce successful companies.