Building a company where AI runs operations, not just assists

  I'm running an experiment.

  A few weeks ago I built 60% of a legal management platform (lex-pro.co) using only Claude Code. Colombian market, real users. I couldn't believe how far I got.

  So I'm stretching it further.

  The goal: a "morning ritual" where I check in once a day to make decisions. AI handles everything else - product development, deployments, customer support, operations.

  Not AI-assisted. AI leading.

  The first problem: AI is blind. Claude can debug code, but only if it can see it. Can help customers, but only if it has context. Right now I'm the bottleneck, copy-pasting everything.

  So step one: build the infrastructure that gives AI eyes.

  That's Brainz Lab (github.com/brainz-lab) - self-hosted observability tools with native MCP support. Logs, errors, APM, feature flags, secrets. Claude can query everything directly.

  Tech stack: Rails 8, PostgreSQL + TimescaleDB, Hotwire. Runs via docker-compose.

  This isn't Lovable or Bolt (type a sentence, get an app). This is trying to answer: can one person + AI actually run a company? Not a demo. A real business.

  Building in public. Happy to discuss the approach, tech decisions, or why this might be crazy.

Sounds like a great idea. I would love to be able to swap technologies, like use a go or node backend, maybe use an entire framework like Echo or NestJS (I’m going for maintainability here, and I feel frameworks provide good guardrails). Bonus if it can be self host able (paid or licensed regardless) and maybe adding in browser rendering so AI can do visual testing. I could be one of the first customers.