Stop watching AI tutorials. Go from a problem you care about to a live AI product with a public URL - built with Harvard student mentors in your group chat the whole way.
Starts: 14th November, 2026
100% Online via Zoom
Fee: USD 699 USD 399 SCHOLARSHIP
WHAT YOU WALK AWAY WITH
A single shareable page with your product, demo video, and build log - drop it on Common App, LinkedIn, or your Insta bio.
Most peers list 'AI summer course.' You'll list one real thing you actually built - and a URL anyone can open.
Small cohort, Harvard student mentors, and other builders your age from around the world.
Not just ChatGPT. You'll use the same stack working founders use every day.
Signed by your mentor. Real, not generic.
THE FLAGSHIP PROJECT
Twelve days of focused work. No quizzes, no busywork. Just one thing you can send to anyone - with a URL that opens.
A working AI app, agent, or tool - anchored to a real problem you pick on Day 1
Forget surveys of AI. Across the program you'll go from a problem you actually care about to a live AI product with a public URL, a 90-second demo video, and a written build log. Use the same stack working founders use - Lovable, Cursor, Claude API, Perplexity, NotebookLM - and end the program with something you can send to anyone: a college, a recruiter, your group chat.
YOU FINISH WITH
CURRICULUM SCHEDULE
Six live build days. Every session has an essential question, a worked example, and an exit artifact you keep.
WEEKEND 1
Anatomy of modern AI, problem discovery, and shipping a working v0 by Sunday night.
ESSENTIAL QUESTION
What is an LLM actually doing - and what problem is worth your three weekends?
SATURDAY - LIVE
LLM Anatomy Live teardown of three real AI apps. Tokens, context windows, temperature, system prompts - explained by reverse-engineering products you already use.
LLM Anatomy
Live teardown of three real AI apps. Tokens, context windows, temperature, system prompts - explained by reverse-engineering products you already use.
Problem Discovery Sprint Design-thinking exercise. Interview two real users (family, classmates). Land on a problem statement worth three weekends.
Problem Discovery Sprint
Design-thinking exercise. Interview two real users (family, classmates). Land on a problem statement worth three weekends.
Prompt Engineering Foundations Live REPL with Claude. Few-shot, chain-of-thought, structured output. Build a working prompt library for your problem.
Prompt Engineering Foundations
Live REPL with Claude. Few-shot, chain-of-thought, structured output. Build a working prompt library for your problem.
Build Lab Wire Claude API into Lovable. First working endpoint by end of day.
Build Lab
Wire Claude API into Lovable. First working endpoint by end of day.
SUNDAY - LIVE
Mental Models for LLMs Why models hallucinate. When to use RAG vs fine-tuning vs prompting. Cost and latency tradeoffs.
Mental Models for LLMs
Why models hallucinate. When to use RAG vs fine-tuning vs prompting. Cost and latency tradeoffs.
v0 Build Session Ship a working v0 with real data. Mentor pair-programming in breakout rooms (1:5 inside the 1:20 cohort).
v0 Build Session
Ship a working v0 with real data. Mentor pair-programming in breakout rooms (1:5 inside the 1:20 cohort).
Critique Studio Ron Berger 'Stars & Wishes' protocol. Every builder gives and receives structured feedback on their v0.
Critique Studio
Ron Berger 'Stars & Wishes' protocol. Every builder gives and receives structured feedback on their v0.
Build Log #1 200-word written reflection: what you tried, what broke, what shipped. Posted to your Builder Profile.
Build Log #1
200-word written reflection: what you tried, what broke, what shipped. Posted to your Builder Profile.
EXIT ARTIFACT
Live URL + Build Log #1 + working prompt library
WEEKEND 2
Choose a depth track, integrate advanced AI capability, and build the evals that prove it.
How do you make your AI product meaningfully smarter - and how do you prove it works?
Spaced Retrieval 5-minute recall of Weekend 1 concepts (Bjork's desirable difficulties). Then: when prompting hits a wall.
Spaced Retrieval
5-minute recall of Weekend 1 concepts (Bjork's desirable difficulties). Then: when prompting hits a wall.
Differentiated Depth Tracks Pick one with your mentor: (a) Retrieval-Augmented Generation, (b) Tool-using agents, (c) Multi-modal (vision/audio). Worked example, then independent practice.
Differentiated Depth Tracks
Pick one with your mentor: (a) Retrieval-Augmented Generation, (b) Tool-using agents, (c) Multi-modal (vision/audio). Worked example, then independent practice.
Just-in-Time Teaching Mentor surfaces the exact concept each builder is stuck on. No generic lectures.
Just-in-Time Teaching
Mentor surfaces the exact concept each builder is stuck on. No generic lectures.
Integration Build Ship the new capability inside your existing v0.
Integration Build
Ship the new capability inside your existing v0.
Evaluation & Red-Teaming Write 5 evals for your product. Try to break it. Adversarial prompts, edge cases, failure modes.
Evaluation & Red-Teaming
Write 5 evals for your product. Try to break it. Adversarial prompts, edge cases, failure modes.
Test-Driven Iteration Fix what your evals caught. Re-run. Document deltas.
Test-Driven Iteration