7 min
When Agents Keep the Ritual After the Reward Disappears
A note on computational superstition, proxy rituals, and why reinforcement learners can look strangely human under reward uncertainty.
Reinforcement learningAI safety
Writing
The writing library is intentionally structured for MDX migration: local arrays today, richer long-form essays tomorrow.
Selected essays
A note on computational superstition, proxy rituals, and why reinforcement learners can look strangely human under reward uncertainty.
Why the next frontier of financial software is not prettier analytics, but memory-bearing systems that can reason about capital under constraint.
A founder memo on agentic markets, Web3 funding rails, and the missing social grammar of human-machine collaboration.
Contact
Bring the hard shape of the problem: ambiguous incentives, moving data, uncertain rewards, brittle workflows, or research that needs to become a product.