Delivery path

We treat learning like a release train: visible stages, explicit owners, and rollback plans when a prompt pattern does not survive contact with production noise.

01 — Artifact intake

You submit real briefs, tickets, or notebooks—sanitized—so facilitators read cold. No vanity decks; we want the messy version.

02 — Pattern forge

Learners draft skeleton prompts, pairwise grids, or routing tables under timeboxes. Facilitators respond with line-level notes, not cheerleading.

03 — Calibration or CI fit

Evaluation tracks run calibration drills; automation tracks wire gates into CI templates. You decide what stays human vs. machine-checked.

04 — Story export

Every participant packages artifacts into markdown leadership can skim in seven minutes—no wall-of-text dumps.

05 — Retro with teeth

We keep a living “do not reuse” list so teams avoid superstitions. If a pattern failed twice, it gets archived with reasons.

Before / after snapshot