A Prompt Template That Makes Qwen 3.8 Behave: Fixing Overthinking and Ephemeral Reasoning
emerybirb · reddit · 2026-09-05
The author shares a system-prompt template (in his jinja setup) that has been working well for Qwen 3.8, aiming for Claude-like behavior with preservethinking=false and context-efficient reasoning summaries.
Core rules:
- Function calls must follow the exact tool-call tag format with required parameters;
- Thinking is ephemeral and discarded after each turn — any conclusion must be written into visible text outside <think> or it is permanently lost;
- Fixed per-turn pattern: short decisive thinking → persistent narration (observations/conclusions/rationale) → immediate tool calls;
- Deferring a tool call or skipping narration counts as an ABORT of the whole turn;
- Thinking is only for resolving genuine tension; read before you think, and if you haven't read something, go read it.
He hopes it helps others hit by common failure modes like overthinking and unstated reasoning conclusions.
More from coding & agent
- 105 Real Bugs Tested: GPT-6.1 Sol Over 10x Cheaper Than GPT-5.6 Sol — PawelHuryn · 2026-10-23
- Dev builds SkiFree-style sled game in one prompt, praises DeepSeek Flash — victormustar · 2026-10-23
- Every ditched personal agents for a shared Claude Managed Agents coworker in one week — every · 2026-10-23
- WorkOS shows how to let AI agents safely sign up for your app with auth.md and ID-JAG — AI Engineer · 2026-10-23
- Devs on Reddit swap tactics for stopping prompt regressions in AI agents — SleepFlimsy6963 · 2026-10-23
- Claude Code + ComfyUI MCP workflow for client work: paid models or open weights? — newaccount47 · 2026-10-23