Learn

• How AI answers

  Follow one prompt from text to reply and see what limits it.

  6 lessons36 min

  1. What is a token, and why is AI priced per token?8 min
  2. Why can't AI count the letters in a word?5 min
  3. Why does a chatbot forget?5 min
  4. Why do answers appear word by word?6 min
  5. Why does the same question give different answers?6 min
  6. What does attention do inside an AI model?6 min

• Use an AI service

  Predict a bill, keep a key private and handle limits.

  5 lessons27 min

  1. Why is my AI app bill higher than I expected?5 min
  2. What is prompt caching, and when does it save money?6 min
  3. Why an AI API key must never go in a web page or app6 min
  4. What 'too many requests' means, and what your app should do5 min
  5. How to stop one user running up your AI bill5 min

• Trust the answers

  Know how an AI app goes wrong, leaks or is tricked, and the plain fix.

  6 lessons36 min

  1. Why does AI make things up?6 min
  2. How do I check my AI's answers with a small test set?6 min
  3. What is prompt injection, and can a prompt stop it?6 min
  4. How does a safety filter decide, and where does it fail?6 min
  5. Can one user's chat leak into another's?6 min
  6. Is it safe to download a model file?6 min

• Inside the chip

  See why moving data, not arithmetic, sets the speed.

  4 lessons22 min

  1. How much memory does an AI model need?6 min
  2. Why a fast AI chip spends most of its time waiting5 min
  3. What does quantization trade away?6 min
  4. How FlashAttention avoids storing the giant score table5 min

• Serving many chats

  See how memory and time are shared between users.

  6 lessons35 min

  1. Why AI servers waste memory, and how paging fixes it6 min
  2. Why does an AI wait a moment before the first word?6 min
  3. How does an AI server answer many people at once?5 min
  4. How can a small model speed up a big one?6 min
  5. Why does a big model need several GPUs?6 min
  6. Why split reading the prompt from writing the reply?6 min

• Plan and measure

  Turn the numbers into a plan and check it by measuring.

  5 lessons27 min

  1. Why is the average speed of an AI service a misleading number?5 min
  2. What do a model server's settings change?5 min
  3. How many GPUs do I need for 1,000 users?5 min
  4. Should I rent a model API or run my own?6 min
  5. Build a tiny model server you can read end to end6 min

• Safety inside the model

  See where a model's refusals live, how they are removed and what still protects an app.

  1 lesson6 min

  1. Can a model's refusals be removed?6 min

Words you meet in a lesson are in the glossary.