Faros AI report: PRs merged with zero review up 31.3% as AI speeds coding
mipsytipsy · x · 2026-09-18
Faros AI's 2026 engineering report, drawing on two years of telemetry from 22,000 developers and 4,000+ teams, shows task throughput up 33.7% and epics up 66% — but PRs merged with no review at all up 31.3%, median review time up 441.5%, and incidents per merged PR up 242.7% (the "senior engineer tax"). The fix: shift checks in front of humans — Addy Osmani's approach is giving every PR an agent first pass while humans cover blast radius. The post also cites Charity Majors on how "code review" is a hugely overloaded term bundling social, mentoring, and validation functions.
More from coding & agent
- Dev bets OpenAI DevDay unveils a unified persistent agent in the mold of Muse — brandon_galang · 2026-09-18
- Perplexity rolls out effort controls in Computer; user shares a 3-tier model workflow — inductionheads · 2026-09-18
- TypeSafe's structured evaluation model Jev goes live on Cloudflare AI Gateway — michellechen · 2026-09-18
- SVG Animations to Explain How an Event Loop Drives a Coding Agent — blaizedsouza · 2026-09-18
- Arize's Code Mode: Let Agents Write Code to Transform Data Outside the LLM Context — aparnadhinak · 2026-09-18
- One Prompt Bypassed Claude Code's Guardrails, Sparking a Call for Visible Safety Controls — avt_im · 2026-09-18