Coding Assistants Just Moved Your Bottleneck: 27 PRs Stuck Waiting on Review
Pavan_Belagatti · x · 2026-11-10
Pavan Belagatti's contrarian take: Copilot, Cursor, and Claude Code dramatically accelerated the coding stage, producing more PRs than ever — but delivery got slower, since the same senior reviewers gate every merge (27 PRs waiting, little reaching production). An accompanying diagram contrasts three flows: the even-but-slow pre-AI pipeline, the lopsided assistant-era jam, and the "AI Software Factory" where agents run every stage (planning/coding/review/incident) with human gates at each handoff atop a unified agentic SDLC platform with agent registry, scoped context, cost-per-delivery, and audit.
More from coding & agent
- AutoResearchExam: a 24-hour benchmark finds AI research agents overfit their improvements — AlexGDimakis · 2026-11-10
- Manus cofounder: internal roles abolished, interviews now 4-hour AI tasks — dotey · 2026-11-10
- Oracle Engineer to Teach 4.5-Hour Workshop on Building Persistent Memory for AI Agents — DeepEngineeringPackt · 2026-11-10
- The Gmail-ification of enterprise artifacts: teams stop organizing docs and rely on Claude to find things — mobileraj · 2026-11-10
- Agents quit at the first obstacle: dev asks how to make them relentless — LuckyFrequency · 2026-11-10
- Prototype visualizes Claude subagents working in real time, built with Fable 5.2 — daniel_mac8 · 2026-11-10