Prefix-reuse FLOPs: a new metric to measure an agent's true compute cost across context edits
RulinShao · x · 2026-11-14
A developer deploying an enterprise context agent found that token savings alone couldn't quantify the value of context edits. Rulin Shao's talk on Context Language Models (CLM) supplied the missing piece: prefix-reuse FLOPs.
- The metric accounts for compute across the entire agent run, including cache reuse and recomputation after context edits.
- Counterintuitive implication: a shorter context can still cost more compute.
- Pairing prefix-reuse FLOPs with task quality gives a concrete way to evaluate whether "beautiful context" actually helps people get better work done.
More from coding & agent
- Hands-on DeepSeek's $0.06/hr ASR API: loses to a locally fine-tuned 1.7B model — AGI Hunt · 2026-11-14
- SAG Wiki: open-source local knowledge base that skips embeddings and vector DBs, works with Claude Code — FellMentKE · 2026-11-14
- Recursive self-improving agent MolRSI lifts wet-lab hit rate 82%, beats ClaudeScience 8-0 — 机器之心 · 2026-11-14
- OpenAI $200 plan drops to $100 on Nov 20; user races to burn 4 banked Codex resets — koltregaskes · 2026-11-14
- Tonbi ships AI Garage plugin giving agents MCP access to his content catalog — Teknium · 2026-11-14
- Google's Modern Web Guidance lifts coding agents' modern web compliance from 56% to 95% — aigclink · 2026-11-14