METR: Expenditure Horizon (blog post)
- Type
- blog
- Venue
- METR
- Year
- 2026
- Source
- blog
- Access
- public
- Language
- en
- Added
- 2026-09-30
- Verified
- 2026-09-30
Summary
METR blog post (2026-07-21) on expenditure horizons, including estimating human expenditure for pull requests with an LLM judge. Shared by Wenhao Chai in the context of agent scaling behavior and the argument that hybrid AI+human setups need careful human-in-the-loop design.
Keywords
agents · expenditure horizon · evaluation · METR · hybrid agents
Topics
agents, expenditure horizon, evaluation, human-AI
Research notes
- Discovery: shared by Wenhao Chai (@wenhaocha1, verified) on 2026-09-30: https://x.com/wenhaocha1/status/2105168366126117102
- Section: estimating human expenditure for PRs with an LLM judge.
- Companion to "When Agents Slow Down" (separately cataloged).