Dev Interrupted Why software factories fail without humans | HumanLayer, Warp & LinearB
62 snips
Sep 1, 2026 
Dexter Horthy, HumanLayer co-founder building collaborative coding-agent infrastructure; Dan Lines, LinearB COO focused on engineering effectiveness; and Aloke Desai, Warp’s founding engineer leading product engineering, debate the human role in software factories. They explore pull-request ownership, safe automation gates, AI’s toughest frontier in incident response, context-layer observability, and how engineers can become builders of the factory.
AI Snips
Chapters
Books
Transcript
Episode notes
Measure Throughput Without Sacrificing Quality
- A software factory succeeds by increasing throughput while maintaining or reducing cost per shipped item.
- Quality must remain part of the equation, because faster automation can otherwise produce unreliable code and frequent rollbacks.
Tie Factory Metrics To Customer Value
- Connect factory metrics to business impact, especially predictable delivery and the amount of customer value delivered.
- Use leading indicators such as effective PR cost, merge rate, rework rate, and assisted PRs to explain progress.
Bad Code Creates A Software Slop Loop
- Optimizing individual workflow stations can hide declining overall throughput and rising long-term costs.
- Software compounds defects: poor code shipped today makes every future change harder until someone fixes it.
