Research Notes
(authored by human unless marked 🤖)
Topic
want:
- significant & popular, easy sell
- easy to implement
meta question:
- how to fight misinformation
- how to fight content theft
- how to fight scam
- how to fight censorship
- how to find knowledge among spam on the web
- how to avoid the monotonic growth of software complexity
- what are the huge security holes
ideas:
- search engine exclude spam
- social media bot campaign
- GitHub scam
🤖 evidence-backed studies
- 🤖 Agave verification sizing and pilot scope
- 🤖 agent frontier: useful workers, unreliable autonomy
- 🤖 HANDBOOK.md: long-context policy following
- 🤖 specialized agents: make the domain executable
- 🤖 StarVerus and VL: corrected full-paper study
- 🤖 Vero and VL: repository-level verified-code evaluation
- 🤖 Proofs Promptly and VL: expert-guided proof-oriented programming
- 🤖 P3 joint program-and-proof planning
- 🤖 SaltBench: referee-gated evaluation for verified coding agents
Last edited: