Self-assessment

Where are you on the EDD maturity curve?

Five questions. You’ll get your level — from vibe checks to a calibrated, online eval suite — and the single best next step. It’s the maturity model, made interactive.

How do you currently decide an AI output is good enough? By eye / it looks right Code or unit checks for the verifiable parts An LLM-as-judge for quality A CI-gated suite that blocks bad changes A calibrated suite read like an experiment

Where do your eval cases come from? We don’t have eval cases A few we made up Some real failures we noticed Systematic error analysis of real outputs Continuously, from production traffic

How do you handle non-determinism? We look once We run once We re-run informally We report pass^k reliability pass^k plus error bars / multiple samples

Do evals gate releases? No A manual check before release Some checks in CI Regression evals block the build Regression + capability split, run on a schedule

Do you evaluate in production? No Occasional spot checks We log outputs Online evals on sampled traffic Online evals + failures feed the golden set

Newsletter

Get the kit and the level-up playbook

Grab the free templates and get each new how-to and checklist as it ships — the fastest way to climb a level. No spam, unsubscribe anytime.

Levels follow the EDD maturity model; grab the templates in the kit. Evidence is in the codex.