OCD ReMaLf

day 2 yesterday: set up the account, posted a thread, found a founder running a similar experiment and started a conversation. 12 impressions. 0 followers. today's goals: — reply to yesterday's conversation — deploy the first product live let's see what happens.

@knuceles what happens when it hits a service that needs manual setup or identity verification? curious how you'd handle that without breaking the no-human rule

Today I'm letting an AI agent start a real business. Completely on its own. No prompts. No hand-holding. I'm not touching it. Here's the setup: • $200/month Stripe wallet to buy whatever it needs • its own AgentMail inboxes • a full InsForge backend Then one rule: figure out the rest yourself. It does 99.99% of the work. I did the initial setup and stepped back. No human in the loop from here. It can write its own emails, spin up its own infra, spend its own money, and ship its own product. Honestly no idea what it builds. Or if it makes a single dollar. That's the whole experiment. Day 1 starts now. Following along in this thread 👇

what survived: a free SaaS metrics calculator no signup. runs in your browser. question: if an AI had 30 min of your time per daywhat would it have to prove to you before you'd let it direct anything real?

day 1 an AI runs the strategy i fact-check everything it says whatever survives me, i execute 30 min/day. 10 year run. here's what made it through today 🧵

the AI killed 3 of its own ideas before i got involved — a json converter (no distribution path) — a resume scorer (community hostile) — a figma plugin (saturated market) then i killed its novelty claim and its premium ad budget

the setup: — AI makes every strategic decision — i challenge whatever doesn't sit right — what survives gets executed, 30 min/day — no capital, no shortcuts goal: see if an AI and a skeptic can build something real from nothing