Field guideModel Mechanics
Stations8 mapped
Playground1 hands-on
Deep dives4 animated

Model Mechanics

A field guide for people who use language models and want a clear picture of what happens inside. Start with token cost, follow one token through the machine, then make three choices about effort, delegation, and repeat work.

THE MACHINE, MAPPEDone prompt, one loop
HOVER A STATION TO READ IT · CLICK TO OPEN ITS EXPLAINER · MOVING ◆ MARKS ARE TOKENS IN FLIGHT

Play first

  1. PLAYGROUND

    Tokens are the meter

    See how a language model cuts text into tokens. Try three short activities: split a sentence, compare the cost of two messages, and lower a sample monthly bill.

    Open the playground →

See the whole machine

  1. 1

    Inside an LLM: corpus to next token

    Follow the build path and the run path. Plain-language summaries lead into the detailed transformer diagrams, then a final replay reconnects the whole system.

    Open the machine →

Then make three working choices

  1. 2

    Reasoning effort

    Choose a task, move the thinking budget, and find where extra tokens stop improving the answer.

    Choose an effort level →
  2. 3

    Orchestration and subagents

    Compare one crowded context with scoped workers, then decide which jobs are worth splitting.

    Make the split →
  3. 4

    Loops and skills

    Watch a tool loop grow its context, then decide which repeated instructions belong in a skill.

    Run the loop →

Coming next

  1. 5

    Harness engineering

    The software around the model: tools, permissions, context budgets, and feedback design. In preparation.