Micro Free Course — 100% Free Learning
All lessons in this module are free to learn. Sign in with Google to save your progress.
LLM Foundations: What They Actually Are
What LLMs are (and aren't); tokens, embeddings, and the transformer in plain language; the pretraining/fine-tuning/prompting spectrum; the 2026 model landscape (closed vs open, sizes, costs); context windows.
Module Content
What LLMs Are (and What They Aren't)
A practical definition. LLMs predict the next token; everything else is engineering on top of that. FIND_VIDEO: search 'what is a large language model how it works' — recommended channel: 3Blue1Brown / Andrej Karpathy. Aim for 11 min or under.
Recap — Tokens, Embeddings, and the Transformer in Plain Terms
The minimum mental model. Skip the math; get the intuition that makes everything downstream easier.
The Pretraining vs Fine-tuning vs Prompting Spectrum
Three ways to get an LLM to do what you want. Most teams pick the wrong one first. FIND_VIDEO: search 'LLM pretraining fine-tuning prompting comparison' — recommended channel: Andrej Karpathy / DeepLearning.AI. Aim for 10 min or under.
Recap — Picking the Right Adaptation Strategy
Pretrain, fine-tune, RAG, or just prompt? Each has different cost, complexity, and effectiveness. Pick deliberately.
How to Compare Models — Closed vs Open, Size, Cost
The 2026 model landscape. Closed (Claude, GPT, Gemini), open (Llama, Mistral, Qwen) — which to pick when. FIND_VIDEO: search 'claude vs gpt vs llama comparison 2026' — recommended channel: AI Engineer / Trelis Research. Aim for 11 min or under.
Recap — The 2026 Model Landscape and How to Choose
Closed vs open, sizes, capabilities, costs. Pick by your task and constraints, not by marketing.
Tokens, Context Windows, and Why They Matter
The fundamental constraint that shapes every LLM application: tokens in, tokens out, with a hard limit. FIND_VIDEO: search 'LLM context window tokens lost in middle' — recommended channel: AI Engineer / Greg Kamradt. Aim for 10 min or under.
Recap — Working Within Token Budgets
Tokens cost money and time. Manage them deliberately; production-grade LLM apps live and die by their token discipline.