Dev runs Claude continuously for two weeks, managing 20 PRs with stacked PRs and jj
audrow · x · 2026-10-17
audrow shares his AI agent workflow: start from a definition of done, break work into cohesive features, and let Claude run essentially continuously for two weeks implementing features while he checks in twice a day to critique. A replyer suggests writing acceptance criteria before the agent reports back to avoid grading leniency. He notes 20 feature PRs in flight, managed via stacked PRs, with the Jujutsu (jj) VCS making the workflow far easier than git.
Related event: Developer shares workflow for running Claude autonomously for two weeks(2 posts)→
More from coding & agent
- Sentry CEO unsure if production sandbox uses e2b or was self-built for GCP — zeeg · 2026-10-17
- Sentry CEO: sandbox workloads run on Vercel, internal services on Northflank — zeeg · 2026-10-17
- Sandbox providers are mostly Firecracker on AWS behind an API, notes developer — nateberkopec · 2026-10-17
- Seroter's reading list: context inertia, agent evals past the vibe check, 3x code shipping — rseroter · 2026-10-17
- Muxin: open-source desktop app to run AI coding agents across all your machines — sull · 2026-10-17
- HeyPCB launches MCP server letting AI agents design PCBs and export KiCad projects — R4nsack · 2026-10-17