Understand AI, as it happens.
Plain-English explanations of how AI actually works, rebuilt as the facts change. Every number is read from a live model index and carries the vendor page it came from and the date it was last checked against it.
The AI field guide: less jargon, more understanding
Find plain-English AI definitions, compare commonly confused ideas, and practice with review cards. Explore the AI field guide, compare terms, or download a cheat sheet.
New: Jev, a model that answers without writing
TypeSafe’s Jev is a System One model. It reads a state, evaluates typed questions against it in parallel, and returns answers with a probability attached to each. It does not write replies, produce code, or explain its reasoning, so the usual comparison against a language model does not apply to it. Most of what was published about it in its launch week does not survive a check against TypeSafe’s own documentation, including the way its name is spelled. Read the lesson, or see its row in the model index and the claims about it that failed.
How to check any of this
- The model index — every figure with the vendor page it was read from, the date it was read, and whether it was still on that page at the last check.
- Claims that did not survive — what was circulating, what the primary source said, and which of those two was wrong.
- What is stale right now — the system reporting its own freshness in public, including the parts that are behind.
10 tracks, in order
Nothing uses a term before the track that teaches it.
- What is this thing? — Start from nothing. Finish knowing what a language model is, what it cannot do, and why.
- Turning the dials — Logits, softmax, temperature, top-p, streaming and stop conditions — the whole control panel, without the maths anxiety.
- Getting good answers — Prompts, examples, thinking budgets, tools and schemas: the controls that decide whether a model hands you something usable or something merely plausible.
- Memory and truth — How a model reaches things it was never trained on — embeddings, retrieval, context and memory.
- What's inside — The architecture without a single equation — from what one parameter is to why the second token is cheaper than the first.
- How models are made — Pretraining, RLHF, LoRA, Chinchilla. The whole path from a pile of text to a model you can call, with the hand-waving removed.
- Cheap and fast — Where the money actually goes: memory, batches, caches, tiers and routing. How to cut a bill by an order of magnitude without changing the product.
- Agents — What the word agent means once the marketing is removed: a loop, some tools, and the permissions you decided to grant.
- Trust and evals — How anyone knows whether any of this works: what an eval is, what a benchmark score hides, and the ways a system fails while its numbers improve.
- Fast, slow, and neither — The System 1 / System 2 story everyone tells about AI, how much of it the evidence actually supports, and where TypeSafe’s Jev sits — a model that answers without writing anything, on neither axis.