> researcher. hacker. founder.
> building systems for the intelligence era.
I build companies, models, and systems around artificial intelligence.
Currently exploring:
AI infrastructure · agents · model routing · AI security
alignment · evaluation · local inference · self-improving systems
Previously built companies. Now mostly interested in what happens when software starts thinking.
Some questions I'm interested in:
→ Can AI systems recursively improve themselves safely? → How fragile is alignment at the weight level? → What does infrastructure look like when models become interchangeable? → Can agent harnesses improve faster than the underlying models? → How do we benchmark agents when benchmarks themselves become training data? → How much intelligence can we run locally?
research > speculation
shipping > slides
experiments > opinions
open systems > walled gardens
The interesting question isn't whether AI becomes more powerful.
It's what we can build once it does.
[████████████████████████████████░░] building...
founder by day
researcher by night
occasionally breaks things to understand them