Three things to do before Module 6: 1. Create your 5 spec files using the mega prompt. Read it before you paste it — then customize it for your product. 2. Audit your own PRD. For every assumption you made — check if your observability plan would actually catch it if you're wrong. 3. Write quality examples for a product that isn't yours. Pick one of the four practice products and write great/bad/edge outputs + must-fail-safely cases.

This is where you submit your work for the three Module 4 activities. Reply to this post with your submissions. What to submit: Activity 1 — The PM Decision Audit Your 4-section diagnosis memo (300–500 words). Include all four sections: what the user expected, what the system did, root cause, and recommended fix. Activity 2 — Design the Invisible Decisions Your answers to all 5 PM decisions for the Spotify "Why This Song?" feature. Be specific — "it should be smart" doesn't count. Activity 3 — The Trade-off Debate Your synthesis paragraph(s). Complete the sentence for each dimension: "Notion AI's approach is better when ___. Gemini's approach is better when ___." How to submit: Make a copy of the google doc in the original post. Add your answers to it. Reply to this post, and include a link to your doc. Peer review (Activity 1 only): After you submit your diagnosis memo, read two other students' submissions and reply to their comment with your peer review. Do you agree with their root cause? Would their fix work? Did they catch something you missed? Drop your submissions below 👇

Before you build AI, you define what "right" looks like. That's a golden set. Your task: Create 10 test cases for a travel itinerary chatbot. Define the user, their message, and exactly what the AI should (and shouldn't) do. Full brief with product context and template linked in the above post. Drop your submission link in the comments 👇

Please share a document with the LLM's name, prompt and the learning summary of session. Please include a visual (optional) Also share in the comments below how would you define "good quality" in this case, and how would you measure success of the "Online classes learning summariser" feature

This week you've got 5 activities that put everything from Module 2 into practice: 1. Fix the Prompt — Take broken prompts and rewrite them using the 5 Elements framework 2. Diagnose the Failure — Figure out why an AI product is giving bad output (hint: it's almost never the model) 3. Design the Context — Map out all 6 context components for a real product scenario 4. Classify the Approach — Decide whether a feature needs a simple prompt, RAG, an agent, or fine-tuning 5. Write a System Prompt — Write a production-quality system prompt from a product brief, then test it live This Google Doc has all 5 activities. Here's what to do: → Make a copy of the doc → Work through the activities → Link your completed copy as a comment on this post