A dev's 4-phase AI workflow: orchestrator agents merge PRs while he watches TV
brandon_galang · x · 2026-08-18
Day 19 of a developer's series overanalyzing his personal AI workflow: four phases — orientation, planning, scoping, execution.
- Orientation & planning: agents catch up on Slack threads and open loops; OpenAI's Codex mobile with GPT Live works great while walking.
- Scoping: adversarial discussion in front of a screen to inspect agent output, narrowing everything into discrete Linear tickets with definitive instructions for sub-agents.
- Execution: babysitting agents in the evening while watching TV. Execution agents run Opus 5 Medium (notoriously slow); a parent orchestrator creates worktrees for sub-agents, delegates work, testing and review, then cuts PRs and merges to main.
He admits phases rarely happen the same day — work stalls in phases 1–3 until he has time to babysit execution, partly because 1Password-based secret management requires physically touching his laptop to approve, a bottleneck he's still streamlining.
More from coding & agent
- Opinion: Bot core split should be IAM, not tasks — willccbb · 2026-08-18
- AI Agent Success Depends on Context Layer, Not Just Models — Pavan_Belagatti · 2026-08-18
- Deploying a Hermes agent on a dedicated phone to automate content tracking — Teknium · 2026-08-18
- Interactive causal network built with Grok 4.6 and vanilla JS: trace, intervene, and reshape systems — techartist_ · 2026-08-18
- Failed AI Agents Consume 40% More Tokens Than Successful Ones — 0xsachi · 2026-08-18
- Swarms Weekly Update: Agent Screener Launches and Expands Hiring — KyeGomezB · 2026-08-18