Standard
Inference
AI, explained plainly.
Calm, clear videos on how AI actually works, for people who don’t work in tech.
Start here
Five videos, in order. Each one answers one question.
- 01
- 02
- 03
- 04
- 05
The series
The foundations.
What a language model is, how it reads, and why it sometimes gets things wrong. Start here.
How to use what you learned.
Practical procedure, built on the foundations: asking better questions and checking the answers.
The maths, for those who want it.
The same ideas, worked through with the numbers. Optional depth, never required.
Coming soonOne idea, under a minute.
- ·Standard Definitions
- ·Can AI actually…?
- ·Skill Tested
- ·Standard Issue
Technical Notes
Longer written notes for readers who want the detail: the maths, the code and the sources.
The softmax function, worked by hand
Turning a list of raw scores into probabilities, one step at a time, with a calculator.
Goes with D1 (coming soon)// Placeholder postInside a tokenizer: byte-pair encoding step by step
How a tokenizer learns its vocabulary by merging the most common pairs of characters.
Goes with Ep2 →// Placeholder postWhy hallucinations aren’t bugs
A model that predicts likely text will sometimes produce likely text that is false.
Goes with Ep3 →// Placeholder postHow we check
Every video is checked and sourced. You can see the working.
- A source for every claim
Each episode page has a public list of its claims, with a source for each one.
- Model behaviour is dated
Anything about how a model behaves names the model and the date we checked it. Models change.
- Mistakes are corrected openly
Corrections are logged and dated on the episode page. See an example →