RESEARCH NOTEBOOK
Notes & writing
- One feedback loop, three kinds of evidence
- Found in the wild: reproducible audits of public agent datasets
- What survived six failed novelty claims
- The detector on code it had never seen: a negative held-out result
- Dynamics blindness: your agent is more like a robot than a chatbot
- Tool descriptions are infrastructure: on Anthropic acquiring Stainless
- The AI tools section, Davey Awards Under/Over Report 2026