1"use strict";(self.webpackChunkwebsite=self.webpackChunkwebsite||[]).push([[2143],{3662:(e,t,a)=>{a.r(t),a.d(t,{assets:()=>l,contentTitle:()=>r,default:()=>h,frontMatter:()=>o,metadata:()=>s,toc:()=>c});var s=a(60679),i=a(74848),n=a(71184);const o={title:"The AI-Native SDLC Playbook Stops at the Database",authors:"a8m",tags:["ai agents","sdlc","migrations","schema-as-code","drift detection","claude code"],slug:"/2026/08/31/ai-native-sdlc-database",description:"Anthropic's AI-Native SDLC playbook turns delivery into a loop of committed artifacts. The loop does not close at the database: why revert-shaped controls fail for schema changes, and the gates that work.",keywords:["ai native sdlc","ai agents","database migrations","schema changes","migration linting","drift detection","schema registry","claude code"],image:"/u/images/ai-native-sdlc/gates.png"},r=void 0,l={authorsImageUrls:[void 0]},c=[];function d(e){const t={a:"a",em:"em",p:"p",strong:"strong",...(0,n.R)(),...e.components};return(0,i.jsxs)(i.Fragment,{children:[(0,i.jsxs)(t.p,{children:["Anthropic recently published ",(0,i.jsx)(t.a,{href:"https://claude.com/blog/the-ai-native-sdlc-playbook",children:"The AI-Native SDLC playbook"}),", and\nit's the most concrete writeup I've seen discussing the effects of cheaper code generation on a delivery process,\nwritten by a company where ",(0,i.jsx)(t.a,{href:"https://claude.com/blog/how-anthropic-secures-its-ai-native-software-development-lifecycle",children:"AI authors 80% of merged code"}),".\nIf you run engineering at any scale, I suggest you read it."]}),"\n",(0,i.jsxs)(t.p,{children:["It consists of six stages from ",(0,i.jsx)(t.em,{children:"Plan"})," to ",(0,i.jsx)(t.em,{children:"Maintain"})," that work in a loop rather than a linear chain of handoffs. Each\nstage commits an artifact that the next stage reads. The chain of commits is the audit trail. Humans stop reading\nevery line and start judging intent and risk."]}),"\n",(0,i.jsxs)(t.p,{children:[(0,i.jsx)(t.a,{href:"https://about.gitlab.com/blog/when-code-is-abundant/",children:"GitLab's CEO"})," responded with the line that will likely outlive\nthe playbook itself: ",(0,i.jsx)(t.em,{children:'"when implementation becomes abundant, trust becomes scarce."'})]}),"\n",(0,i.jsxs)(t.p,{children:["Code is abundant now. Your data is not. An agent can regenerate your code in an afternoon, but it cannot regenerate the\nproduction state your company accumulated over years. The playbook's recovery model quietly assumes otherwise, which\nleads to our claim: ",(0,i.jsx)(t.strong,{children:"the loop does not close at the database"}),"."]}),"\n",(0,i.jsx)(t.p,{children:"The database is the one stateful component in your infrastructure. It's where the history lives, and you cannot roll\nback history by redeploying an older version of it."})]})}function h(e={}){const{wrapper:t}={...(0,n.R)(),...e.components};return t?(0,i.jsx)(t,{...e,children:(0,i.jsx)(d,{...e})}):d(e)}},60679:e=>{e.exports=JSON.parse('{"permalink":"/blog/2026/08/31/ai-native-sdlc-database","source":"@site/blog/2026-08-31-ai-native-sdlc-database.mdx","title":"The AI-Native SDLC Playbook Stops at the Database","description":"Anthropic\'s AI-Native SDLC playbook turns delivery into a loop of committed artifacts. The loop does not close at the database: why revert-shaped controls fail for schema changes, and the gates that work.","date":"2026-08-31T00:00:00.000Z","tags":[{"inline":true,"label":"ai agents","permalink":"/blog/tags/ai-agents"},{"inline":true,"label":"sdlc","permalink":"/blog/tags/sdlc"},{"inline":true,"label":"migrations","permalink":"/blog/tags/migrations"},{"inline":true,"label":"schema-as-code","permalink":"/blog/tags/schema-as-code"},{"inline":true,"label":"drift detection","permalink":"/blog/tags/drift-detection"},{"inline":true,"label":"claude code","permalink":"/blog/tags/claude-code"}],"readingTime":15.08,"hasTruncateMarker":true,"authors":[{"name":"Ariel Mashraki","title":"Building Atlas","url":"https://github.com/a8m","twitter":"arielmashraki","imageURL":"https://avatars.githubusercontent.com/u/7413593?v=4","key":"a8m","page":null}],"frontMatter":{"title":"The AI-Native SDLC Playbook Stops at the Database","authors":"a8m","tags":["ai agents","sdlc","migrations","schema-as-code","drift detection","claude code"],"slug":"/2026/08/31/ai-native-sdlc-database","description":"Anthropic\'s AI-Native SDLC playbook turns delivery into a loop of committed artifacts. The loop does not close at the database: why revert-shaped controls fail for schema changes, and the gates that work.","keywords":["ai native sdlc","ai agents","database migrations","schema changes","migration linting","drift detection","schema registry","claude code"],"image":"/u/images/ai-native-sdlc/gates.png"},"unlisted":false,"prevItem":{"title":"DACPAC for PostgreSQL: Where Atlas Fills the Gap","permalink":"/blog/2026/09/23/dacpac-for-postgres"},"nextItem":{"title":"Case Study: How Health in Tech Maintains Databases and Security Compliance with Atlas","permalink":"/blog/2026/08/11/case-study-health-in-tech"}}')}}]);
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.