Take one startup — Cook & Bake Academy Singapore — from an idea to a fully operating business in two days, using ChatGPT, ChatGPT Work and Codex together.
| Course detail | Information |
|---|---|
| Course code | C695 |
| Programme | Non-WSQ |
| Duration | 2 days / 15 instructional hours (9:30am – 5:30pm) |
| Registration | View course details and register |
Learners play the AI-assisted team behind Cook & Bake Academy, a fictitious cooking and baking school with two campuses and 20 hands-on courses (based on the Codex demo site). Each lab starts from what the previous one produced:
- Validate — market research in Chat and ChatGPT Work, published as an HTML report with
@Sites. - Build — Codex plans and builds the website, a sign-up form for every course, and a fully static SQLite RAG course assistant (HTML/CSS/JavaScript, no backend) that deploys to GitHub Pages or Sites.
- Operate — ChatGPT Work prices the courses, puts the term on Google Calendar, handles enrolment and Gmail reminders, and runs customer service.
- Grow — marketing know-how becomes custom skills;
$imagegenmakes the flyer; a multi-agent studio writes social posts and YouTube packages; the open-house campaign closes the loop.
Along the way learners use the Codex features that make an agent dependable: AGENTS.md, /plan, /goal, plugins, @Computer Use, skills from skills.sh, custom skills, hooks, custom agents and scheduled tasks — and evaluate every result against evidence before it goes out.
By the end of the course, learners will be able to:
- Analyse a range of agentic AI applications and identify their strengths, limitations and industry applicability.
- Establish the correlation between agentic AI algorithm design and efficiency, and evaluate process improvement.
- Assess methods for evaluating the effectiveness of agentic AI applications.
- Evaluate agentic AI applications for performance effectiveness through comparative analysis.
| Day | Topic | Labs |
|---|---|---|
| 1 | 1. Fundamentals: Chat, Work and Codex — evolution of AI engineering, harness engineering and the agent loop, OpenAI products and GPT-6 models, desktop app and plugins, the 7-step workflow, /plan, AGENTS.md, publishing |
1–4 |
| 1 | 2. Tools and the SQLite RAG Assistant — SQLite FTS5 knowledge base in the browser, /goal, bring-your-own-key ChatGPT mode, red-teaming, @Computer Use QA |
5–8 |
| 1 | 3. Skills and Hooks — SKILL.md vs AGENTS.md, installing skills from skills.sh, $skill-creator, hooks, a CI deploy gate and scheduled tasks |
9–11 |
| 2 | 4. Running the Academy with ChatGPT Work — pricing and break-even, @Calendar, @Drive, @Gmail enrolment and reminders, a customer-service skill |
12–15 |
| 2 | 5. Digital Marketing — marketing skills, the $imagegen flyer, custom agents for posts and video, lead magnet, newsletter and campaign analysis |
16–18 |
Each lab has its own folder with the scenario and context, a step-by-step README (Markdown and PDF), copy-paste prompts (Markdown and PDF), starter assets, a solution state where the lab produces code, and an evidence checklist. Start with the scenario and the labs index.
- Validate the Idea with Chat and Work
- Plan and Build the Site with /plan
- Project Rules and a Sign-up Form for Every Course
- Publish the Site: GitHub Pages and Sites
- Turn the Brochures into a SQLite Knowledge Base
- Build the Course Assistant and Drive It with /goal
- Add ChatGPT Mode, Then Try to Break It
- QA the Whole Site with @Computer Use
- Install Community Skills from skills.sh
- Create Custom Codex Skills
- Guard Rails That Run Themselves
- Price the Courses and Find Break-even
- Put the Term on the Calendar
- Enrol the Learners and Automate Reminders
- Customer Service with a ChatGPT Skill
- Marketing Know-how as Skills, Then the Flyer
- A Multi-Agent Content Studio
- Lead Magnet, Newsletter and the Open-House Campaign
- Courseware v12.1 in courseware/:
- Learner Guide (Markdown, v12.1) — concepts and the full step-by-step procedure for every lab
- Scenario and labs index
- 18 self-contained lab folders; Lab 11's solution holds the complete verified Cook & Bake site
You need the ChatGPT desktop app (download) with Codex, a GitHub account, and a Gmail account. All business data is synthetic; every learner email address in the fixtures resolves to your own inbox through plus-addressing.
This public repository contains the courseware, learner-safe guidance, synthetic data and lab assets. Only the current courseware version is published. Source references, build tooling, archived versions, credentials, .env files and QA artifacts are intentionally excluded. The API keys in Lab 11 (sk-proj-TEST…) are deliberate fakes used to prove the secrets hook works.
Tertiary Infotech Academy Pte. Ltd.
UEN: 201200696W