Skip to main content
DATUM
Enter

Practice the call before it counts.

One hand at the Table a day, twelve short lessons, and toys that show how AI works by letting you move it. No account, and nothing here costs anything to get wrong.

Today's TableTuesday 27 October 2026
Five calls, the same hand for everyone today.Today leans on: Own the call.

Your lesson path

0 / 12
  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
Start with lesson 1: The context window

Pick any lesson. Each one lets you try the idea for yourself.

Opening your Table history…

This week: Keep a person in itDecide which steps a tool takes alone and which wait for a look.

This week's thread

A lesson, a shift and a project that lean on Keep a person in it. Take any, or none.

Twelve lessons

Each one is a live page that shows one idea inside an AI. Start with work.

Using AI at work0 of 7 explored

  1. The context windowAI can only see a set amount of text at once. In a long chat, the oldest parts quietly fall out of view.
  2. RetrievalBefore answering, the system looks up the most relevant bits of your documents and hands them to the AI along with your question.
  3. Prompt injectionText inside an email, web page or file can trick an AI into treating it as an order, as if you had written it.
  4. EvalsBefore you swap one prompt or tool for another, test both on the same set of real examples and score them the same way.
  5. Picking the next tokenAI writes one chunk at a time by guessing what comes next, with a dial for how safe or adventurous each guess is.
  6. TokensAI does not read letters or whole words. It reads chunks, and the chunks it knows came from what it read most.
  7. Precision and recallAny system that flags things has to choose between catching everything and only flagging what is truly a match.

How it is built0 of 5 explored

  1. EmbeddingsAI turns words into lists of numbers, so words that mean similar things end up pointing the same way.
  2. AttentionWhen reading a sentence, each word looks back at the other words and decides which ones matter most for its meaning.
  3. Underfitting and overfittingA model can be too simple to see the pattern, or so eager that it memorizes every quirk and mistake as if it mattered.
  4. Hold some data backTo find out how good a model really is, test it on examples it never saw while it was learning.
  5. Gradient descentTraining is many small adjustments, each one making the model a little less wrong, with a setting for how big each step is.

Words you will meet

Shifts

Work a day with an AI tool, one decision at a time.

Seven working days, a few minutes eachA support lead, an analyst, a recruiter, a ward manager, a developer and a shop owner. What you skip at nine comes back at three.

Toys

Move an AI and watch what it does.