no-mistakes(review): Fix F1 model rename and F2 128K compaction math in contracts

2026-07-23 09:04:51 +00:00
parent 1c44bf1259
commit e6c52bf071
2 changed files with 3 additions and 3 deletions
@@ -59,7 +59,7 @@ prefill time at 532 tok/s. Fix context first, routing second.
 - **Route by task**: qwen for code/standard queries; gemma for
  compression/auxiliary; strix-moe for compression tasks.
 - **Compress aggressively**: threshold at 40% (not 65%) — a 128K window should
-  compact at 85K, not 166K. Target 15% tail (not 30%).
+  compact at 51K, not 85K. Target 15% tail (not 30%).
 - **Cache everything repeated**: system prompts, skill docs, AGENTS.md — these
  never change between turns. Single-digit cache hit rate is unacceptable.
 - **Lower context ceiling**: 128K window is the stable ceiling for agent conversations.