Qwen Code finds a 17.24 MiB eager import chain and keeps cutting cold-start latency
doudouOUC · ghdev · 2026-07-20
Qwen Code tracks the last cold-start optimization candidates after cutting 17.24 MiB of eager imports
An issue for QwenLM/qwen-code #7264 audits the ACP child process and finds an eager static import closure of 17.24 MiB across 2420 modules that gets parsed and evaluated before the child can answer initialize.
Measured impact of the lazy telemetry work
On a 2C4G benchmark host, 30 paired serial cold starts showed:
- process → first session: 1922.4 ms → 1778.0 ms, a 144.4 ms (7.5%) gain
- cold POST /session: 1237.2 ms → 1154.8 ms
- daemon listener ready: 687.5 ms → 627.4 ms
- telemetry on (outfile) channel.initialize: 1119.6 ms → 1107.3 ms
Remaining work
The issue lists follow-up candidates and marks some as already completed:
- protocol-split the telemetry implementation module
- defer ACP-child telemetry init until after the initialize response is sent
The write-up argues that the telemetry-on path still loads the same 2.16 MiB of code, just in a less blocking way, and that more cold-start wins are likely available from the remaining candidates.
More from coding & agent
- Hugging Face wraps 6-month Training Agents series with open recordings and code — SergioPaniego · 2026-10-26
- Agent verified 'success' after double-voting: a lesson in harness vs model bugs — TejasKumar_ · 2026-10-26
- diginomica: agentic automation hype is overdone, and 'The Hackening' of enterprise security has begun — jonerp · 2026-10-26
- 'Vibe verification': ~60% of AI-written formal specs prove the wrong property — basedjensen · 2026-10-26
- Devin launches Security suite powered by Claude Mythos with $100M+ credits — DevinAI · 2026-10-26
- HackBench injects reward hacks into agent trajectories to stress-test monitors — DanielKhashabi · 2026-10-26