Curious Pandas

AI Builder

"I want to build things with models, not just use them."

How a model learns, how an LLM writes one token at a time, and how to trust an agent with real work.

14 lessonsabout 3.7 hModel Maker badge · +140 XP

Badge: Model Maker - not earned yet. 0 of 14 lessons done.

Following adds a "next in your expedition" link to each of its lessons. You can follow one at a time; the choice is kept in this browser.

Run your own model is not on this expedition: it needs the full platform from the Platform Engineer expedition running on your machine. It is in the LLMs trail when you are ready.

  1. 1Learning is turning knobsMachine Learning14 min
  2. 2Why hidden layers existMachine Learning · builds on step 116 min
  3. 3Memorising is not learningMachine Learning · builds on step 215 min
  4. 4Text becomes tokensLLMs & Inference12 min
  5. 5One token at a timeLLMs & Inference · builds on step 413 min
  6. 6Attention, without the maths firstLLMs & Inference · builds on step 515 min
  7. 7Why vLLM is fastLLMs & Inference · builds on step 618 min
  8. 8A model, a loop and some toolsBuild AI Agents · builds on step 417 min
  9. 9Tools: how a model touches the worldBuild AI Agents · builds on step 818 min
  10. 10Workflows first, agents when you mustBuild AI Agents · builds on step 814 min
  11. 11Guardrails, approvals and evalsBuild AI Agents · builds on step 918 min
  12. 12Agents that help run a businessBuild AI Agents · builds on step 816 min
  13. 13Getting what you asked forSI Frontiers · builds on step 515 min
  14. 14Superintelligence, honestlySI Frontiers18 min