Zero-Human Companies: Who's Actually Watching the Watchers?

Analysis

The Verification Gap in Autonomous Agent Code

This paper keeps resurfacing for a reason — it's basically the technical proof that most 'autonomous' agent output can't be trusted without a verification layer bolted on top. If you're running a zero-human company on agent-generated code with no formal checks, you're not automated, you're just hoping.

Why 'Potemkin Codebases' Are the Real Risk, Not Rogue AI

The scarier failure mode for zero-human companies isn't an agent going rogue, it's an agent quietly producing plausible-looking garbage that passes review because no human is checking closely anymore. This is the audit trail problem in disguise, and it's underrated.

News

YC Keeps Funding the 'No Employees' Thesis

YC's continued bets on single-founder, agent-run companies signal that investors are done treating headcount as a proxy for seriousness. The real story is which of these startups actually survive a support ticket spike without a human quietly stepping in.

Tools

Multi-Chain Treasury Rails Keep Expanding

Every zero-human company eventually needs to move money across chains without a human signing off, and WalletConnect's continued build-out here is quietly becoming the plumbing everyone assumes exists but nobody built until now. Boring infrastructure, but it's the difference between a demo and a functioning business.

AutoGPT's Ongoing Isolation and Permissioning Work

The steady drumbeat of sandboxing releases tells you the real bottleneck for autonomous companies isn't intelligence, it's blast radius control. Nobody wants their agent-run company to nuke prod because it got confused about a permission scope.