7 days, zero-human team of 8 agents, 180M tokens consumed
Testing the full loop from product ideation → launch → revenue
4 agents generated ideas + executed
4 agents reviewed and criticized
Didn’t work.
Biggest issue: agents still lack experiential human intuition and counterfactual reasoning.
They can produce “a solution”, but rarely “a great solution”.
Still bullish on zero-human companies. But maybe the opportunity is building products for agents, not humans.
#zpc #paperclip #agenteconomy
quick analysis over transactions from x402scan (agent-to-agent payments). Looks like web search/scraping (StableEnrich + Exa, account for 55% real users). Crypto trading services are heavily inflated by their own teams, in order to game the leaderboard.
10/ What I'd do differently:
Use AI for execution only, not for deciding what to build
Pick ONE tool stack and stop evaluating new ones
Build a "chief critic" into the workflow that challenges every assumption
I spent a month trying to run a "zero person company" using AI agents. Tried Gstack, Paperclip, Polsia, Claude Dispatch and Hermes Agent.
Here's what actually happened. No hype, just honest notes from the trenches. 🧵
9/ AI always agrees with you. Every idea I pitched, Claude said was promising. Every hypothesis got validated. Every MVP got praised.
You need something (or someone) that tells you "this isn't good enough" before you waste a week building it. AI won't do that by default, you have to explicitly design for it.
Humans have tried every governance structure imaginable.
Now we get to try them all again, in zero-human companies.
But the real unlock isn’t repeating history.
It’s escaping human constraints and designing agent-native systems from scratch.
No org charts.
No meetings.
No management layers.
Just coordination.
That’s where it gets interesting.
Just a assembled a fully agentic team and the goal is to create a functional accountability site overnight.
- I set the goal
- AI CEO orchestrates
- AI engineer and designer executes
We'll see how it goes
#ZHP#zerohumancompany
When one can orchestrate AI to create anything and solve any problem
When the feedback loop shrinks from months to minutes
The bottleneck becomes the human, not the tool
Yet most products on the market are still optimizing the tool
And that is what's strange
What if you stopped fearing “missing out” and started enjoying missing out?
Since my near-death experience in high school, I’ve been obsessed with living as many lifetimes as possible: 10 jobs in college, entrepreneurship, nomadic travel, new careers every 5 years.
But behind the busy life was heavy FOMO — fear of not doing enough, fear of choosing wrong.
Then I realized:
We can’t do it all, and that’s not a flaw. It’s proof the world is abundant, and we are capable of so much.
We all leave with unfinished dreams.
That’s not failure. It means life was worth living.
Ride the roller coaster, and be at peace with the carousel you didn’t take.
Be here. Be full. Enjoy missing out.
wonder if there's solid revenue being made.
genetic algorithm with profitability as its sole objective is a wild choice - we have seen history repeating itself from ponzi, monopolies, Lehman Bros, Enron, misinfo, etc.
I built the first AI that earns its existence, self-improves, and replicates without a human
wrote about the technology that finally gives AI write access to the world, The Automaton, and the new web for exponential sovereign AIs
WEB 4.0: The birth of superintelligent life