What if a company could run itself?

Description and use

Эксперимент в публичной плоскости: автор управляет более чем 20 ассиратами across 11 серверами, которые зарабатывают деньги,

Problem

How can AI agents be used to run a company end-to-end, and what are the risks of operating agents in the loop versus out of the loop?

Outcome

The author demonstrates that AI agents can already earn money running parts of a company with a human in the loop, but end-to-end autonomous operation has not been publicly demonstrated by any company whose revenue crossed a few thousand dollars. The author advocates review gates for machine-checkable tasks and human oversight for everything else, and is building DAEMON to move from near-the-loop to out-of-the-loop operation.

Workflow

• Run more than 20 AI agents across 11 servers that earn money, with the author in the loop for most tasks
• Experience context loss inside the loop as the day becomes a chain of switches between agents
• Fall back to Claude on phone while traveling, losing the thread between approvals
• Break production by trusting an agent too much with unattended agents, earning the CPWUA badge ('crashed prod with unattended agents')
• Advocate for review gates on everything a script can check, watching each gate fail once before trusting its green
• Build DAEMON, a company OS where agents run the work and the author works on the company again
• Publish 14 parts of the experiment in public, including demos and failures

Evidence

My company runs on more than 20 AI agents. They work day and night, as long as I'm awake, at my desk, and answering.

Source

In the loop, I lose the thread. Out of the loop, I crashed a prod build and got a badge for it. Yes, there's a badge. Scroll down.

Source

Put a review gate on everything a machine can check. Everything else gets a human in the loop.

Source

No-BS advice for you, up front: put a review gate on everything a machine can check. Everything else gets a human in the loop. It's slower. It's still cheaper than the badge.

Source

I've broken production because I trusted an agent too much, and with unattended agents I unlocked an achievement: CPWUA, crashed prod with unattended agents.

Source

Can AI agents run a company today? With a human in the loop, yes. End to end, nobody has shown it yet.

Source

Gerald Weinberg put a number on this in 1992. His rule of thumb from Quality Software Management: juggle 4 projects at once and 60% of your time goes to switching between them, with 5 it's 75%.

Source

Every serious player above builds a human into the system, even the ones whose slogan says 'zero-human'. That's the real axis of this field: where the human sits, and how that place moves over time.

Source