About iDiscoverAI

The problem

There is no shortage of information about AI. There is a shortage of information you can act on.

Ask a simple question — what does this model cost, how big is its context window, can I run it on my own hardware — and you get three answers from three sites, none of them dated, at least one of them wrong. The vendor's own page is marketing. The directories are spreadsheets with logos. The blog posts were true in March.

That is not a content problem. It is a structural one, and it has two halves.

The first half: everything has one home

Most AI sites are organised as a category tree, and the tree always collapses. "AI agents" ends up as a section under Learn, a section under Technology, a section under Tools, and a top-level category of its own — four pages competing for one subject, and no answer to the question of where a specific agent actually lives.

Here, every single thing is an entity with exactly one canonical URL. A model is a model. The product wrapped around it is a tool. The idea underneath it is a concept. Those rules are mechanical, they are published, and the database enforces them — slugs are unique across every kind, so nothing can be filed in two places even by accident.

Everything else on the site — the comparisons, the collections, the learn tracks, the glossary — is a view over those entities. A view links to entities. It never restates them.

The second half: everything is dated

In most fields, a fact from 2019 is merely old. In this one it is wrong. Context windows change, prices halve, models get retired.

So every page here carries the date a human last checked it, and that date has teeth: after a kind-specific interval — 90 days for models, two years for concepts — an unverified entry automatically drops out of our search index and shows a warning to readers. Nobody has to remember. The clock does it.

You can see the consequence in the footer: the total number of entries, and the smaller number that are currently verified. We would rather show you the gap than hide it.

Who this is for

People who have to decide something today. An engineer choosing a model, a founder working out what an agent can be trusted with, a student trying to understand what a transformer actually does without reading the paper first.

How it is funded

Not by placement. There is no paid inclusion, no sponsored entries, no affiliate-driven ordering, and no way to buy a position on a list. The editorial policy says exactly what we do and do not accept, and the methodology says how entries are checked.

What is missing

A lot. This is a reference site being built in public, and the honest state of it is visible on every page: entries that have not cleared the publish gate render with a note saying so. If something you need is missing or wrong, tell us — it goes straight onto the editorial queue, and no account is required.