Former Google AI leader Jeff Dean has released a 1-hour lecture covering all of AI engineering. The scope runs from LLMs from scratch → prompts → agent teams → graphs, a full course from "0% to 100%."
From LLMs to Agent Teams in 52 Minutes: A Map of the Lecture
| Segment | Time | Content |
|---|---|---|
| 0% | 0:00 | LLMs from scratch: the start of what made Google |
| 30% | 17:22 | How to actually use AI models |
| 65% | 30:03 | Prompt engineering |
| 100% | 52:35 | Scaling to agent teams and graphs |
Jeff Dean's Hour: Why It Shows the Shape of AI Engineering
1. Jeff Dean is an "architect of AI infrastructure." It matters in itself that the person who built TPUs and led Google's large-scale AI is teaching from the beginner level.
2. Not just how to use it, but the whole picture. Starting from how LLMs work, it goes all the way to "teams" and "graphs" where multiple agents collaborate, drawing a map of AI engineering on one sheet.
3. For beginners and practitioners alike. It runs from 0% to 100%, so you can find your place and start from there.
Watch From the Start or Pick Segments: How to Use the 52 Minutes
If you want to move from "using" AI to "building" with it, this lecture is a good entry point. The prompt (30:03) and agent scaling (52:35) segments are especially useful at work right away. It is an hour long, but the timeline works like a syllabus, so it is easy to pick just the parts you need.
Related: X post · YC interview