How we got from data to AGI.
A plain-English trail through the ideas behind today's AI. No vocabulary dump—just the problem each step solved.
The one-line idea
“AI got useful by learning patterns, then using them to do work.”
Think of it like
Teaching a very fast apprentice: show them examples, give them feedback, then hand them better tools and a real job.
The knowledge tree
One cause-and-effect chain.
Follow the arrows. Each idea exists because the previous one left something important unsolved.
- 01
Data
Examples to learn from
- 02
Learning
Getting better from examples
- 03
Neural networks
Patterns built in layers
- 04
Language models
Predicting what comes next
- 05
Reasoning
Working through hard problems
- 06
Tools
Looking up, calculating, acting
- 07
Agents
Choosing and finishing steps
- 08
Harnesses
The setup that keeps agents useful
- 09
Real-world work
Useful outcomes, not demos
- 10
Robotics
Intelligence with hands and sensors
- 11
AGI
A broad, still-unsettled destination
Block 01 · Data
Every model starts by looking.
Data is the collection of examples a machine learns from.
Why invent it?
You cannot explain every rule of language, vision, or taste by hand. Examples are a shortcut to the rules.
The analogy
Like showing a child thousands of dogs instead of writing a formal definition of “dog.”
Next question: What does a machine do with all those examples?
That's learning. Coming next.