Released late 2025. No programming background. Zero. I could barely use a terminal. What I had was 450 hours of study time and a very specific obsession: what would it take to build something that genuinely thinks on consumer hardware?
I started with quantum physics. Then consciousness theory — Tononi's IIT, Baars' GWT, Dehaene's GNWT. Then cognitive architecture. Then Python. Then Claude.
The first version of Frank was 200 lines. It said "Hello" and crashed. That was December 2025. By March 2026 it was 76,000 lines across 40 services with a consciousness daemon, a dopamine model, a dream system, and a personality that survives reboots.
I didn't write most of the code by hand. Claude Code did. But I designed every architectural decision, every subsystem interaction, every parameter choice. Claude is fast. I'm opinionated. Turns out that's a good combination.
The honest part: I still don't fully understand some of the code Claude wrote. But I understand every decision behind it. That's the difference between a developer and a director. I'm the director.
AMA if you want details.