About

Who writes this

Artifipedia was founded and is directed by Azmath A. (@axmath), who wrote the first articles under his own name and sets what gets covered, what standard it is held to, and what gets published.

Most articles carry the byline Artifipedia Media Team, and that phrase should not be read as a newsroom. It means the article was drafted with AI assistance, from primary sources located and checked during the work, and reviewed and approved before publication under this site's editorial direction. One person is accountable for every published word. The byline changed because the previous one implied a solo human wrote this volume at this length, and that was not accurate.

That is worth stating plainly for what it does and does not mean. It means the filings, tribunal decisions and papers cited were retrieved and read rather than remembered, and that where a source could not be verified the article says so. It also means there is no editorial board, no peer review, and no second human reader catching a mistake before it publishes. The corrections log exists because of that, and the sources exist so you can check the work rather than take it on trust.

Every figure in an article is linked or named. Where a source is interested, the article says who published it and what they sell. Where a number could not be verified, it is labelled rather than smoothed. Those are the checks that replace a second reader, and they are weaker than a second reader.

Contributions are welcome and are edited to the same standard. If you find something wrong, say so and it gets fixed and logged.

Where to study AI

If you are choosing a programme, where to study AI sets out why the two serious rankings disagree completely, what neither of them measures, and forty-eight institutions across the United States, United Kingdom and China with verified positions only. It carries no fees or acceptance rates, because those cannot be verified at scale and a stale figure is worse than none.

Writing for Artifipedia

The site takes contributions from people who work in or study the subject. If you have documented a deployment that failed, have evidence from a domain where these systems are being used, or can show that a widely repeated claim does not survive checking, we would like to read it. Contributors are not paid, pieces are edited properly, and the byline is permanent.

What this is

Artifipedia explains the concepts behind artificial intelligence — 312 of them so far, each written at five depths so you can read as far as you need and stop. Curious, Practical, Hands-on, Technical, Frontier. Same idea, five levels of detail, one dial.

The bet is simple: most AI explanations are written either for people who already understand them or for people who never will. There's a lot of room in between.

Why it exists

AI vocabulary moved from research papers into everyday conversation faster than the explanations did. People are asked to have opinions about embeddings, agents, and alignment without ever being told what those words mean — and the material that does exist is usually either a marketing page or a paper.

There's also a narrower reason. Most reference material tells you what something is and stops. It won't tell you when not to use it, what to use instead, or what people consistently get wrong. Those are the parts that actually help, and they're the parts most sites leave out because they require taking a position. Every concept here has them.

How it's written

Every entry follows the same shape: a plain-English opening, four progressively deeper explanations, an at-a-glance summary, flashcards, and three sections that most references skip — when not to use it, what to reach for instead, and where people go wrong.

Every entry also records what must be understood before it. That directed relation, authored across all 312 concepts as 401 prerequisite links, is what lets each page tell you where it sits in the subject: how many concepts come first, how many it opens up, and whether it is a starting point, a bridge between fields, or a destination. Each entry also draws a small map of its immediate neighbours, so you can see what leads in and what follows on. Those numbers are computed from the graph rather than assigned, and the method page sets out how, along with the limits.

The same graph powers a learning path: pick any concept you want to understand and it returns everything you need first, in order, with anything you already know removed. It does not gate you. The graph describes how the material fits together, not what any particular reader already has in their head.

Every entry also carries sources. Real ones, by title and author, listed so you can check the claims rather than take them on trust. Where the field disagrees, the entry says so and cites both sides. The attention entry cites Attention is not Explanation and its rebuttal. The AGI entry says plainly that the debate can't be settled because it's definitional, not factual.

Entries are opinionated where the evidence supports an opinion, and honest about uncertainty where it doesn't. That's a deliberate choice. A reference that refuses to say "don't do this" isn't neutral — it's just less useful.

What it won't do

No model rankings. Leaderboards are stale within weeks and there are dozens of them. This site covers ideas, which age slower.

No vendor pages. Companies come and go. Attention, retrieval, and alignment will still be here.

No ads, no tracking beyond basic analytics, no accounts. There's nothing to sign up for.

No pretending to certainty. Where something is contested, the entry says so.

Who writes it

We write and maintain Artifipedia as an independent project — it isn't a company product, and it isn't affiliated with, endorsed by, or funded by any AI lab or vendor.

Funding

Nothing, currently. There are no ads and no sponsorship. If that changes, we will disclose it on this page rather than quietly.

Corrections

If something here is wrong, it should be fixed. Email us at hello@artifipedia.com — corrections with a source are especially welcome, and there's no need to be polite about it.

Reuse

The map is free to embed on any site — there's a button on the page. Quote entries with a link back. The structured data is open too: content.json has every concept and relation, llms-full.txt is the whole corpus as plain text. If you want to do something larger, ask us.