A practical AI course for technical people

Understand AI.
Apply it. Own it

Most courses teach the tool. ExplAIn teaches the system underneath.

Listen to the first episode free See what’s inside

One email unlocks it, and it plays anywhere on the site. The waitlist gets the opening date and the founding price.

The transformation

From AI builder to AI owner

Builders make AI work. Owners know why it works, and where it breaks.

Builder

  • Swaps the model when an answer comes back wrong
  • Rewrites the prompt and hopes it holds
  • Calls the whole product “the AI”
  • Ships on a demo that worked once

Owner

  • Names the part that failed, then fixes that part
  • Decides what enters the context, and in what order
  • Knows where the model ends and the system begins
  • Ships on a set of real cases that has to keep passing
Join the waitlist

One email. No payment now, and the founding price is held for you.

The system map

The model generates. The system does

Nine parts decide whether AI works, and ten modules teach them. Point at a part to see which module it belongs to, and what that module covers.

What it sees

05

Context

What the model sees right now

 06

Knowledge

Documents and retrieval

 06

Memory and state

What persists between turns

The model Generates the next token, and nothing else

02

What it reads

Tokens, embeddings, attention

 03

How it learned

Training shapes it, inference uses it

 04

How it answers

Sampling, and why it hallucinates

What it does

08

Orchestration

Steps and loops around calls

 07

Tools and actions

What it can do in the world

 01

Interface

Where people meet it

Around all of it

07

Guardrails

What it may see and do

 09

Evaluation

How we know it works

The ground it stands on

00

How we got here

Centuries of maths, two winters, and one architecture

The mechanisms

Two things everyone gets wrong

Module 03 · Probability

The capital of France is

  • Paris71%
  • located9%
  • a6%
  • now4%
  • Lyon2%
Illustrative. The model ranks every possible next token, then sampling picks one. That is why fluent text can still be wrong.

Module 07 · Agents

Goal Model decides Tool acts Observe
An agent is not a personality. It is this loop, with the model choosing the next step each time round.

The podcast

The first episode is free

Portrait of Simone Vale, the guest on episode one

Episode 00 · with Simone Vale

How we got here

Almost none of the maths is new. How old it really is, why the field stalled twice, and what actually changed.

Listen free — one email

The episode is yours for an email address. Enter it once and it plays here and anywhere else on the site, on this browser.

Want to hear what’s next? Join the waitlist

The curriculum

Ten modules.
One complete map

One mental model each, one sentence you keep, and the failures it explains.

00The historyFree
“Almost none of the maths is new”
Open module 01The map
“The model generates. The system does”
Open module 02Inside the model
“Context is the model’s working reality”
Open module 03Learning
“Training changes the model. Inference uses it”
Open module 04Probability
“Fluency is not the same as truth”
Open module 05Context
“Construct the world the model reasons inside”
Open module 06Knowledge
“Information, state and behavior are different problems”
Open module 07Tools
“The model requests. Another system acts”
Open module 08Agents
“An agent is a loop with model-selected decisions”
Open module 09Evals
“A demo is not a system”
Open module
Read the full course syllabus

Opens soon

The founding price goes to the list

One email now. You get the launch date, a single-use founding code, and the first chapter before it is public.

Join the waitlist

No payment now. Unsubscribe anytime.

From the blog

Articles

All articles