Claude Code burns ~15-20% of your tokens on 3-word status lines; Bun preload turns it off
arpit_bhayani · x · 2026-10-31
A gist by Flamerix AI reveals that Claude Code fires a model call roughly every 30 seconds per background agent, sending the agent's full transcript just to generate a 3-5 word status line like "Reading runAgent.ts" — quietly consuming 15-20% of your tokens.
- Tested on Linux with Claude Code 2.1.257–2.1.312 (macOS reportedly working), no binary modification required
- The fix: a Bun preload script that disables agentsummary at startup
- Shared by arpitbhayani (130K followers), confirming the mechanism
For anyone running background agents, this is an invisible cost that can now be eliminated immediately.
Related event: Claude Code Status Bar Secretly Burns Up to 26% of Tokens(4 posts)→
More from coding & agent
- Curtis Yarvin proposes the "self-driving computer" where every app builds itself — ZeroStateReflex · 2026-10-31
- Non-dev vibe-codes a bike showroom's full management app, now asked for a voice agent — SaiTarunRonanki · 2026-10-31
- Designing agent evals: breaking timestamp ties with revisions without order dependence — jylusdev · 2026-10-31
- Self-Credit: self-generated credit assignment boosts long-horizon LLM agents by up to 8.5 points — gregd_nlp · 2026-10-31
- herdr hits YC and 32K followers in 5 months with durable terminal runtime for coding agents — Scobleizer · 2026-10-31
- Your AI coding agent is not your coworker yet: who owns the architecture? — Obaid_BLM · 2026-10-31