Teach with Vibe

Everything you need to run Vibe as a workshop, an intensive, a cohort or an in-house enterprise programme. The curriculum is the same in every format; only the depth and length change.

Choose a format

Format Length Content Best for
Workshop Half day (3.5 hours) Module 00 plus the highlights of Modules 05 and 06, using the starter repo Founders, business owners, community events. Your entry point.
Intensive 2 days Gate 1 of the Production Readiness Checklist (the 38 "do now" items) applied to each attendee's own app Founders with an app already live. Attendees leave with the Vibe Builder (L1) credential.
Cohort 4 weeks per term One term of the programme at a time, 10 to 12 hours a week Product people, analysts, junior engineers
Enterprise 16 weeks, or a custom subset The full programme, with the AGENTS.md template and the readiness checklist adopted as company policy Banks, insurers, corporates rolling out AI coding tools

Run every session on the same four beats

Every module is written for this rhythm. Keep to it and the class runs itself.

  1. Learn (15 to 20 minutes). Open with the real incident, then the concept. The incident creates the need; the concept answers it.
  2. Build (60 to 90 minutes). Learners direct their AI tool using the prompt pack at the end of the module. You walk the room as the tech lead, asking "how do you know that works?"
  3. Break (30 to 45 minutes). Learners attack their own build, or swap with a partner, using the lab steps. The labs are designed so that the AI's first attempt fails here. This is where the learning happens.
  4. Prove (15 minutes). Learners tick the Definition of Done in their Vibe account and attach evidence: test output, a scan report, a screenshot. No evidence, no pass.

Explain less, break more. Most of the value is in beat three.

The starter repo: Ledgerly

Ledgerly is a small multi-tenant invoicing app, delivered as a specification and a skeleton. Learners ask their AI to build each feature. Fourteen security acceptance tests define what "safe" means: access control, injection, secrets, XSS, rate limiting, password storage, N+1 queries, timeouts, webhooks, race conditions, SSRF, prompt injection, error handling and CORS.

Every test fails until its feature is built safely, and the AI's first attempt usually fails at least one of them. That moment is the lesson. Learners run npm run score and watch their score climb from 0 / 14 to 14 / 14. It gives every class the same starting point and makes progress visible on the projector.

Download the starter repo

Rules for learners: use your AI assistant to build every feature, read every diff before accepting it, and never edit the tests.

Your first workshop, step by step

  1. Two weeks before. Send attendees the setup check: Node.js 22 or later, an AI coding tool they can use (Claude Code, Cursor or similar), and the starter repo installed with npm install.
  2. Room setup. One projector for the workshop deck, one for the live scoreboard. Good Wi-Fi matters more than anything else.
  3. On the day. Follow the workshop deck. It contains the run of show with timings and speaker notes (press N).
  4. Afterwards. Send the readiness checklist link and an invitation to the intensive. Ask for one sentence of feedback while it is fresh.

Running an intensive on attendees' own apps

Day one covers security and data (Modules 03, 05, 06). Day two covers reliability and shipping (Modules 07, 08, 11, 12, 13). Each attendee works through Gate 1 of the readiness checklist against their own product, recording evidence in their Vibe account. Award the Vibe Builder (L1) credential only when every Gate 1 item has evidence.

Two practical warnings: real apps contain real secrets, so start with a secrets scan and rotate anything leaked before you share screens; and agree up front that nobody runs fixes against production during the session.

Running the enterprise programme

Assessment and certification

Level Credential Requirement
L1 Vibe Builder Terms 1 and 2 passed; app passes Gate 1 of the checklist
L2 Vibe Engineer Terms 1 to 3 passed; app survives the Term 3 chaos lab
L3 Enterprise Vibe Architect All terms plus the capstone, defended before a panel

Keep the automatic fail rules. A credential is only worth something if it is hard to fake: a secret in git history, cross-tenant data access, or an agent taking an irreversible action without approval means the module is not passed.

Facilitation tips