"One thing harder than reading AI code is reading AI tests." Mikuel from Safe Intelligence argues spec driven development leaves a loop open: you have a markdown spec, but how do you know the product actually behaves that way? His answer is Cucumber, nearly forgotten and suddenly useful again. Executable, human readable BDD scenarios connect directly to PRDs and critical user journeys and close the gap between what the spec says and what the tests verify.

The rest of the talk is enforcement. ADRs capture not just what the rules are but why; agents rejected at commit time get linked back to the document and iterate. Module import linting makes N+1 queries structurally impossible: rendering templates cannot touch the database, E2E tests cannot import any module that could. His sessions run 20 to 50 context compacts. The agent stays on track because the rules live in git hooks and CI, not in the prompt.

Speaker info:
- https://cz.linkedin.com/in/michal-cichra-61188a84

AI Engineer

A conference talk exploring how to capture and enforce architectural and product decisions for both human developers and AI agents. Covers Architecture Decision Records (ADRs) for documenting why code is shaped a certain way, Product Requirements Documents (PRDs) for lightweight feature context, and Behavior-Driven Development (BDD) with Cucumber as an executable, human-readable specification layer. Also discusses design systems for consistent UI and a feedback loop using git hooks, CI, linters, and architecture checks to keep both humans and AI agents aligned with established rules across long autonomous sessions.

BDD, ADR, PRD, WTF: Capturing Decisions for Humans and AI Alike — Michal Cichra, Safe Intelligence