Developer: I spend more time steering AI coding agents than actually coding
No_Computer_1013 · reddit · 2026-10-18
A Reddit developer notes that Claude Code/Codex sometimes finishes an afternoon's work in 20 minutes, but other days are spent repeatedly correcting agents: "don't touch that file," "that helper already exists," "read the CodeRabbit comments first." He asks production-codebase users what their last heavy-steering task was, and whether agents.md, rules, skills, or hooks actually made the fixes stick past a week.
More from coding & agent
- Agent memory's blind spot: lessons learned in one task silently become rules for the next — blaizedsouza · 2026-10-18
- Multi-agent coding is 'the greatest slot machine casinos could never match' — theteknosaur · 2026-10-18
- AI makes it too easy to build every feature — a spec keeps shipping disciplined — audrow · 2026-10-18
- Dev uses Claude to mine 1,270 Apple Notes from 10 years into a private archive of recurring ideas — inexistences · 2026-10-18
- Dev releases LIT Power Tools: a 44-tool local MCP server plus LM Studio memory plugin — LunderIT · 2026-10-18
- Agent self-audit cuts monthly bookkeeping from 3-4 hours to under an hour — RachelVT42 · 2026-10-18