What the check measures
AI-assisted development does one thing: it produces more code, faster. Whether that code is good or bad is decided by what happens after the suggestion lands — whether the type system, the tests and the pipeline let it through or say no. The AI readiness check measures exactly those guardrails, not how much AI you use.
Twelve questions, three per part:
- Guardrails in the code — how strictly typed the codebase is, whether there are typed contracts between frontend, backend and services, and whether lint and static analysis block in the pipeline.
- The test gate — what happens when a test fails, whether critical flows are covered by end-to-end tests, and how quickly a developer hears back from the pipeline.
- Reproducibility — whether a complete environment can be raised from code, how rollback works, and whether every developer runs the same development environment.
- Ways of working & ownership — whether all code is reviewed before merge regardless of who or what wrote it, whether there are guidelines for AI tools, and who owns architecture decisions when AI suggests something.
Each answer scores zero to three points, 36 at most. From 26 points the gate holds: AI suggestions become hypotheses that have to prove themselves. Between 13 and 25 it leaks — there is speed to be had, but some suggestions pass without proving themselves. Below 13 there is no gate, and AI accelerates the mistakes too. The result points out the part where you lose the most, because that is where starting pays off.
Everything is computed in your browser. Nothing is sent, nothing is stored, and you can redo the check as often as you like. If you want to act on the result, Developer Experience is the service about exactly these guardrails, and the AGENTS.md generator is a concrete first step.