Dev reads 40k words of AI output daily, shares 5-step workflow to cut it down
Wise-Reflection-3701 · reddit · 2026-09-05
A 7-year developer realized Claude Code had him reading 12k lines (40k words) of AI output per day — a novel's worth — and shares his 5-step fix:
- Make the model write less: a CLAUDE.md/AGENTS.md block mandating a strict report format (changes, non-changes, decisions needed; no recaps; 150-word cap) cut his daily volume by 1/3.
- Compress what he didn't control: a fixed prompt that squashes any long spec into a 30-second read while preserving identifiers and error messages.
- Read diffs in order: file list first, deleted lines before added (models delete what they don't understand), then hardcoded values, new deps, network calls, and test assertions.
- Listen instead of read: uses macOS Spoken Content to hear specs while testing branches; he built a small tool (Kairo) for this.
- Clean before pasting: rewrites AI text into plain first-person messages and edits a few words before anything leaves his machine under his name.
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