The AI Experiment Log

Documenting what happens when one person + AI tries to build real companies. Building in public, sharing the wins and lessons along the way.

Jan 2025 Complete

Launched personal site with Claude Code

Built this entire site using Claude Code as my AI pair programmer. From design to deployment, Claude handled the heavy lifting while I made the decisions. First real test of the AI-first development approach.

Jan 2025 In Progress

Brainz Lab documentation with Mintlify

Building comprehensive documentation for Brainz Lab using Mintlify. Claude is writing docs, I'm reviewing and refining. Learning how to scale documentation with AI assistance.

Dec 2024 Complete

Built LexPro MVP 100% with Claude

Shipped a complete legal management MVP with Claude doing all the code generation. Rails backend, React frontend, database design, API endpoints. Proof that AI can handle real product development, not just boilerplate.

Dec 2024 Ongoing

Started the experiment

Left Picap to test a crazy hypothesis: one founder + AI can build and ship real companies faster than traditional teams. Created Brainz LLC and started working 100% with Claude Code as my development partner.

By the numbers

~40
Days running
Since starting Brainz LLC
2
Products launched
LexPro & Brainz Lab
100%
Generated with Claude
Zero code written by hand

What I'm learning

Claude is a developer, not a copilot. The difference is subtle but important. Claude doesn't need me to write code. Claude needs me to make decisions: what to build, how to architecture it, which tradeoffs to accept.

Code generation was never the bottleneck. I can write code. Everyone can write code. The hard part is deciding what to build and why. Claude takes the former off my plate so I can focus on the latter.

Quality matters from day one. There's no room for "we'll refactor later." Since Claude generates code, every line needs to be production-ready. This forces better architecture thinking upfront.

Shipping is the real skill. Building is the easy part now. Getting users, iterating based on feedback, handling support—that's where the actual work is. AI handles the code, I handle the business.

What's next

Scaling both products. Getting real users on LexPro. Making Brainz Lab the standard for AI-native infrastructure. If this works, it proves that the future of software is one person + great AI. Let's see if I'm right.