Turning off preserve_thinking cuts Qwen coding sessions' context use by half with little loss
YetAnotherAnonymoose · reddit · 2026-11-03
A Reddit user reports that disabling preservethinking on Qwen 3.8 massively extends coding-agent sessions: over 50% of the KV cache was thinking traces, and the model works fine seeing only what it did, not how it reasoned. Key trick: instruct via AGENTS.md to write gotchas and hard-won findings into a memory file. Result: little capability loss, permanent notes on pitfalls, and 2/3 fewer context compactions.
More from coding & agent
- AI Slop Addons Flood World of Warcraft, Forcing Warcraft Logs to Rework Its API — 404 Media · 2026-11-03
- Hot take: the old-school monolithic web form was perfect for one-shotting agents — aarthir · 2026-11-03
- 20-year language dream handed to Opus: 200 pages of notes turned into an executable plan — pbaylies · 2026-11-03
- Paradev AI sidebar now agent-driven, surfacing diffs, files, browser and mobile app views — massimosgrelli · 2026-11-03
- DeepSeek Harness fever: two WeChat groups fill to 200 members in 10 minutes — vista8 · 2026-11-03
- Send an article link to your Claude dot and it builds a responsibilities page — craigsdennis · 2026-11-03