LLM Fundamentals & Your First Chatbot
You can now build a domain-specific chatbot from scratch and put it in front of a real user.
3-hour live session · 90 min theory · 90 min hands-on · plus one exercise
The 3-hour live session
Theory grounds you. Hands-on lands it. Both happen in the same Saturday session.
Theory
Open the magic box. By the end of this session, ChatGPT and Claude stop being mysterious - they're fancy prediction machines, and you'll know exactly what they can and cannot do. Tokens, embeddings, context windows, and the five hard limits of LLMs that govern every product decision from here on out. The session that makes 'AI hype' impossible to take seriously again.
Hands-on
Build a real chatbot end-to-end. Watch it forget your name (it has no memory). Fix that. Give it a personality through a system prompt. Deploy it to a live URL. Send the link to someone outside the cohort. The first time you watch a stranger talk to something you shipped - that's Project 1 complete.
Your exercise for the week
Build a domain-specific version of the chatbot for your own product or role. Make it answer five real questions your users would ask. Bring the responses to next session.
The moment that lands
NanoGPT trained live in Cohort 1 on Shakespeare. 85M parameters, loss descending on terminal from ~4.0 to ~1.4 during the session. The output is plausible Shakespeare-style text with errors. Hallucination as a structural fact, not a bug to fix.
What I'll keep saying
- “There are only two core concepts in LLMs - tokens and embeddings. Everything else revolves around that.”
- “At the end of the day, these are just fancy prediction machines.”
- “This session is your cheat code to stop treating AI like magic and start using it like the fancy calculator it really is.”
Want this week in your real life?
DM me on WhatsApp or book a 30-minute call. The cohort is 12 to 18 PMs and I personally vet every applicant.