Interactive widgets
Widgets you can run on their own, and embed anywhere
Each one is a single interactive stage from the course, on its own URL: predict what the mechanism will do, then watch it. No account, no code. Every page carries an iframe snippet, so a lecture, a wiki or a blog post can carry the widget itself instead of a screenshot.
Predict the next word
Language needs context.
Pairs with the explainer: Next-token prediction
Query, Key, Value
Each token emits a Query, a Key, and a Value; match Q to K by dot product.
Pairs with the explainer: Query, Key, Value
Context window & KV cache
Fixed context length; the KV cache reuses past keys/values for speed.
Pairs with the explainer: KV cache
Mixture of Experts
A router sends each token to a few expert FFNs; sparsity decouples cost.
Pairs with the explainer: Mixture of Experts
Distillation: a small model learns from a big one
A small student model imitates a big teacher, far cheaper to run.
Pairs with the explainer: Distillation
RAG: retrieval as a callback to similarity
Embed the query, find the nearest chunks by dot product, paste into context.
Pairs with the explainer: RAG vs fine-tuning
From the internet to your answer
Trained once on filtered internet text (guess, miss, nudge, repeat), then frozen; live answers come from tools feeding the frozen model.
Pairs with the explainer: Training vs inference
The lethal trifecta
Danger spikes when an agent has all three: private data, untrusted content, and a way out.
Pairs with the explainer: Prompt injection
How embedding works
Every widget page has an Embed section with a ready iframe. The frame loads a chrome-less version of the stage, runs entirely in the reader's browser, and links back to the full lesson. It needs no API key and no attribution beyond the link that is already inside it.
Teaching a class or maintaining a reading list? Link the widget page rather than the frame, so your readers get the written explainer and the lesson around it.
Where the rest of the course is
These are eight stages out of ninety-nine. The full course walks from a word is a vector to attention, the transformer block, RAG, agents and the hardware underneath, one predict-then-reveal lesson at a time.
One idea at a time, in your inbox
New lessons and explainers, written the way these pages are. Now and then, not daily, and never a sales sequence.
We email you a confirmation link first. Unsubscribe in one click, any time. Privacy.