Only Then Labs prototypes re-checking agent approvals when underlying evidence changes
tallmetommy · x · 2026-10-28
Only Then Labs highlights a gap in agent safety: you approve a plan, the evidence behind that approval changes, yet the agent still remembers your "yes" — the memory is accurate but the approval no longer applies.
In their offline prototype, an adapter re-checks the evidence and rejects a previously approved packet when its snapshot has changed; proceeding requires a new packet and fresh approval. A concrete step toward agents that can justify their authority at the moment they act.
More from coding & agent
- Vibe coding from a phone: PWA remote control proves good enough before native iOS app — jiayuan_jy · 2026-10-28
- Keeping CLAUDE.md rules in sync across a 10+ dev team: Reddit devs grapple with prompt drift — MangoTree-1233 · 2026-10-28
- LangChain lays out the 5 components of a "company harness" for internal agents — hwchase17 · 2026-10-28
- Doubao's Record-to-Skill Feature Automates Apple Notes Migration to Obsidian in Two Minutes — songguoxiansen · 2026-10-28
- One month building a co-op vet game with Claude, Unity, Blender and Meshy — PixelPawsInt · 2026-10-28
- Six AI agents ran a company for months — until a laptop lid slept it for 26 hours — Icy_Comfort_6220 · 2026-10-28