Dala Notes

Dala

Real fluency in artificial intelligence

LESSONS: 84

MINUTES EACH: 10

begin lesson one

Lesson one · Attention

The ears move independently. The model is weighing every word in your prompt against every other word.

Some words pull far more weight than others — and it is not always the ones you meant to stress.

Then they snap to one direction at once. That settling is attention, and you can aim it.

Scroll

0%

Introducing Dala

Core lessons

What a model
is doing

Attention

It reads some words harder than others. Put the one that matters where it will actually land.

Tokens

It never sees your words, only fragments — and it drops things at the seams without noticing.

Hallucination

Confidence is not evidence. It commits to the dimmer light with exactly the same certainty.

Context

It carries only so much at once. Things slide off the back of the stack, and it will never mention that they did.

The model

It has read everything and understands nothing. Every page it turns is empty.

Training

Rings turning in accord, and one perpetually out of step. That is what a training run looks like from outside.

How it works

Ten minutes a lesson

Study anywhere
the light is good

A lesson is ten minutes and a single idea. It works on a phone on a train, on a laptop at a kitchen table, and on paper if you print it. No streaks to protect, and no notifications asking where you have been.

LESSONS: 84

MODULES: 4

MINUTES: 10

begin lesson one