A curated learning path. Free. About thirty hours total if you watch it all. Start at "Start here" and walk down the tree as your curiosity directs.
flowchart TB
S[Start here] --> V1[What is an LLM?]
V1 --> V2[Inside the Transformer]
V2 --> V3[Building GPT from scratch]
V1 --> V4[Prompt engineering]
V4 --> V5[RAG from scratch]
V5 --> V6[Agents and tools]
V6 --> V7[Multi-agent systems]
V7 --> V8[MCP and the agentic platform]
V3 --> V9[Fine-tuning and LoRA]
V8 --> V10[Frontier news and analysis]
1. 3Blue1Brown — But what is a GPT? An introduction to transformers — https://www.youtube.com/watch?v=wjZofJX0v4M The best visual introduction to how Transformers work. No prerequisites. Grant Sanderson's signature animation.
2. 3Blue1Brown — Attention in transformers, visually explained — https://www.youtube.com/watch?v=eMlx5fFNoYc A natural follow-up. Exactly what self-attention is, drawn on a chalkboard the way you'd want it drawn.
3. Andrej Karpathy — Intro to Large Language Models (1 hour) — https://www.youtube.com/watch?v=zjkBMFhNj_g A one-hour state-of-the-field talk for beginners-to-intermediates. If you only watch one video from this list, make it this one.
4. Stephen Wolfram — What is ChatGPT doing and why does it work? — https://www.youtube.com/watch?v=flXrLGPY3SU A three-hour deep dive with Lex Fridman for a more mechanistic, ground-up explanation.
5. Andrej Karpathy — Let's build GPT from scratch, in code, spelled out — https://www.youtube.com/watch?v=kCc8FmEb1nY The most important coding tutorial in the AI world. Over two hours. You write a Transformer in PyTorch line by line.
6. Andrej Karpathy — Let's build the GPT Tokenizer — https://www.youtube.com/watch?v=zduSFxRajkE The boring-until-it-bites-you subject of BPE tokenization, taught with the same patient rigor.
7. Andrej Karpathy — Neural Networks: Zero to Hero (playlist) — https://www.youtube.com/playlist?list=PLAqhIrjkxbuWI23v9cThsA9GvCAUhRvKZ Eight videos, roughly twenty hours. You will emerge on the other side as a more dangerous engineer.
8. Andrew Ng + OpenAI — ChatGPT Prompt Engineering for Developers — https://www.deeplearning.ai/short-courses/chatgpt-prompt-engineering-for-developers/ Short course, ninety minutes total. The foundational "how to prompt" habits every engineer should have.
9. LangChain — RAG from Scratch (playlist) — https://www.youtube.com/playlist?list=PLfaIDFEXuae2LXbO1_PKyVJiQ23ZztA0x Fourteen short videos that walk you through every lever in a RAG system: indexing, query translation, routing, retrieval, generation, evaluation.
10. DeepLearning.AI — Advanced Retrieval for AI with Chroma — https://www.deeplearning.ai/short-courses/advanced-retrieval-for-ai/ Short and practical; covers query expansion, cross-encoders, and evaluation.
11. Anthropic — Building effective agents (engineering blog + associated talks) Not a single video, but the foundational framing for production agents. Searchable on Anthropic's YouTube and blog.
12. LangChain — What's next for agents — https://www.youtube.com/@LangChain Harrison Chase's talk series. Evolving every few months.
13. Building MCP servers (various) — YouTube search MCP server tutorial
MCP is still new enough that the best videos are community tutorials rather than polished courses. Pick any recent one with good views and watch at 1.5×.
14. AI Engineer Summit / World's Fair talks — https://www.youtube.com/@aiDotEngineer The practitioners' conference. Shorter, denser talks than academic conferences. Search for talks on your specific stack.
15. Jeremy Howard — A Hackers' Guide to Language Models — https://www.youtube.com/watch?v=jkrNMKz9pWU Ninety minutes of practical, opinionated, run-your-own-model wisdom from fast.ai's founder.
16. Hugging Face — The Alignment Handbook walkthroughs — https://www.youtube.com/@HuggingFace Short, code-first videos on SFT, DPO, LoRA, and eval.
17. Full Stack Deep Learning — LLM Bootcamp — https://fullstackdeeplearning.com/llm-bootcamp/ A full day of recorded lectures on prompting, evaluation, UX, and LLMOps.
18. Yannic Kilcher — https://www.youtube.com/@YannicKilcher Goes deep on new papers with healthy skepticism and drawn-on-screen explanations.
19. AI Coffee Break with Letitia — https://www.youtube.com/@AICoffeeBreak Shorter, well-produced paper overviews for the commute.
20. The AI Epiphany (Aleksa Gordić) — https://www.youtube.com/@TheAIEpiphany Long-form paper reads; excellent on diffusion and multimodal models.
21. AI Explained — https://www.youtube.com/@aiexplained-official Measured, skeptical analysis of frontier releases. A good antidote to hype.
22. Dwarkesh Podcast — https://www.youtube.com/@DwarkeshPatel Long-form interviews with lab leaders and researchers. Denser than most podcasts; often remarkable.
23. Latent Space — https://www.youtube.com/@LatentSpacePod Practitioner-focused. Episodes on specific products, teams, and techniques.
24. Lex Fridman Podcast — https://www.youtube.com/@lexfridman Occasionally excellent interviews with Altman, Amodei, Hassabis, LeCun, Karpathy, and others.
25. StatQuest with Josh Starmer — https://www.youtube.com/@statquest If you want the ML fundamentals (linear regression, backprop, neural nets) taught slowly and cheerfully.
26. Two Minute Papers — https://www.youtube.com/@TwoMinutePapers Short, enthusiastic paper recaps. Fun between deeper watches.
27. The Wall Street Journal — How ChatGPT works — Search YouTube for "WSJ ChatGPT behind the scenes" interviews. Good for sharing with less-technical stakeholders.
gantt
title Four-week video plan (about 3 hrs/week)
dateFormat YYYY-MM-DD
axisFormat %b %d
section Week 1
What is a GPT (1h): done, 2026-05-04, 2d
Intro to LLMs (1h): done, 2026-05-05, 2d
Prompt engineering course (2h): active, 2026-05-07, 3d
section Week 2
Build GPT from scratch (2h): 2026-05-11, 3d
Build the tokenizer (2h): 2026-05-13, 3d
section Week 3
RAG from scratch playlist (3h): 2026-05-18, 5d
section Week 4
Hackers' guide to LMs (1.5h): 2026-05-25, 2d
Agents + MCP talks (1.5h): 2026-05-27, 2d
AI Engineer highlights (1h): 2026-05-29, 1d
At the end of four weeks you will have watched the most important AI content on the open internet.
Back to index.