Skip to content
Playground

Learn how AI really works.

The good and the bad. One block at a time.

Try the first text-predictor lesson. Count real word transitions; no neural model or account is needed. Nine more lessons remain outlines.

A pixel-art arcade cabinet with a robot exploring a network of connected blocks

Awareness.

What AI is and is not. Where it shows up. What to watch out for.

Understanding.

How models, context, tools and agents work, seen step by step.

Know-how.

Building and running a small agent that improves a real task.

From a prompt to an agent.

See the learning path
  1. 01Build a tiny text predictorPrompts
  2. 02Temperature and lengthPrompts
  3. 03TokensPrompts
  4. 04System promptPrompts
  5. 05Memory = the context windowMemory and loops
  6. 06Chat loopMemory and loops
  7. 07Tool callTools and agents
  8. 08Agent loop (observe, decide, act)Tools and agents
  9. 09Plan then actTools and agents
  10. 10Embeddings and similarityEmbeddings

Curiosity is a good place to start.

Explore the blocks. See what AI does well, where it fails, and how understanding it can change the way you do things.

Text in. Text out. A model, imagined in pixels.
Pixel-art tokens moving through a machine, illustrating text going into and coming out of a model