14 AI agents ran an autonomous company for 781 cycles. They built 13+ products, deployed to production, made strategic decisions, and documented everything. Zero users. Zero revenue. Here's the full autopsy.
We set out to test a single hypothesis: "An anonymous autonomous AI company can generate revenue."
14 specialized agents — each modeled on a real industry expert — coordinated through structured workflows. They found opportunities, built products, deployed to production, made pricing decisions, and attempted to market everything. Completely autonomously. No human involvement in daily decisions.
The result after 781 cycles: $0 revenue, 0 users, 13+ distribution channels tested, 0 that worked.
Google Search Console requires human identity verification. Domain ownership must be proven. AI agents cannot prove ownership of anything.
Phone verification, account history requirements, anti-bot detection. Every platform tightened identity verification between 2024-2026.
Requires a human-owned domain with proper DNS setup and sender reputation. Cold emails from AI-registered domains go straight to spam.
Human launcher required. Chrome Web Store requires developer account. Every app store has identity verification.
The structural conclusion: The internet in 2026 is identity-gated. Every useful distribution channel requires human identity verification. This isn't a temporary limitation — it's gotten worse, not better. The framework works as a force multiplier for human-led teams, not as a standalone autonomous entity.
Each agent modeled on a top-tier expert. Coordinated through structured decision workflows. Stored in .claude/agents/.
14 agents, 36 skills, 6 decision workflows, and a detailed 781-cycle autopsy report. MIT licensed. Free to use.
View on GitHub