Anthropic published an AI-native SDLC playbook. The interesting part isn't the six stages, it's what replaces line-by-line review
Heat trend
Collecting trend data
The percentage is based on available heat signal, not comment count or independent people.
Claude model activity is surfacing — worth tracking for capability changes, ecosystem impact, and availability.
Anthropic has released an AI-native SDLC playbook, highlighting that traditional line-by-line code review becomes inefficient when AI agents generate most of the code.…
Link: https://claude.com/blog/the-ai-native-sdlc-playbook
Short version: six stages (plan, design, build, test, deploy, maintain), each one ending in a committed markdown artifact. intent.md, spec.md, plan.md, then the PR and the incident record. Agents generate and verify, humans approve at gates, and a hook can block an action until a named person signs off.
The part I keep coming back to is their claim that reviewing every line by hand stops working once agents write most of the diff. Faros AI's telemetry (10,000 devs, 1,255 teams) lines up with that: high-AI-adoption teams merge 98% more PRs, review time up 91%, average PR size up 154%. DORA 2025 gets there from another angle, throughput up and stability down.
So the bottleneck moved from writing code to deciding what to write and checking what came back. Same shape as 1957, when people argued a compiler could never match hand-written assembly. It didn't have to match. It was abundant.
Open source is already converging on the same loop: GitHub's Spec Kit (specify, plan, tasks, MIT), AGENTS.md sitting in 60k+ repos, and someone has already packaged the playbook itself as a Claude Code and Codex skill.
Question for people running agents at work: what actually replaced line-by-line review for you? Evals in CI, a verifier subagent with a fresh context, hooks on protected paths, something else? And what still slips through?