Yaz A. Caleb.

Was born as "Yağız Erkam Çelebi". Grew up in Ankara, half-Albanian, self-taught English and Computer Science, and finished high school 3 years early to be able to pursue building full-time. Most of what I know came from open courses and documentation: Harvard CS, MIT lectures, YC Startup School, and a lot of debugging at 3 a.m. I like systems work, product architecture, and writing clearly about what I'm building.

At 14 we built Clade AI on my front porch - a hardware-native assistant that ran off-device. We shipped prototypes, learned the hard problems of voice, latency, and inference at the edge, then shut it down when the team collapsed. Around the same time I became one of the first civilian testers of GPT-2 and DALL·E 1, which pulled me deep into generative models and human-machine interfaces.

That same year I gave a TED-Ed talk on Terms and Conditions. Since then I've kept a steady rhythm of publishing essays that pressure-test ideas in public. I train, read, and ship. I prefer small, composable tools over platforms that try to do everything.

Today I'm mostly building agent-native security and developer infrastructure. The computer now writes more code than humans can audit, so I'm focused on AI sovereignty, trust, observability, and safe full-autonomy. I live mostly on the road, work remotely across Europe and Asia, and look for high-agency collaborators. If something matters, I build it, measure it, and keep what survives contact with reality.