Would coding agents do fine with just bash? Benchmarking without read/write/edit tools
lucasmeijer · x · 2026-09-21
The author proposes a coding-agent eval idea: strip away the dedicated read/write/edit tools and leave models with just bash to see if that's enough. Bash alone can handle reading, writing, searching and editing files, so dedicated tools may be unnecessary framework complexity. A comparison benchmark someone should actually run.
More from coding & agent
- Dev running local models with Codex suspects kv cache reuse is underperforming — TheZachMueller · 2026-09-21
- Jev classifier model opens to everyone; early users run browser agents at $0.001 per task — PrajwalTomar_ · 2026-09-21
- New Grok Bot x Jev template gives your agents a fast calibrated classifier — eptwts · 2026-09-21
- Nautilo Guide: Stand Up a Self-Hosted AI Assistant Locally with Just Docker — Dan_Jeffries1 · 2026-09-21
- Agent Arena Analyzed 20,840 Coding-Agent Traces: 69% of Complaints Are Broken Code — arena · 2026-09-21
- DeepMind, Meta and Amazon release 135-page roadmap redefining AI agents — mdancho84 · 2026-09-21