Method
039 records · 38 builders
How records get made
There are already enough places that list AI tools. This archive exists for the question those places cannot answer: could I actually build this, and what would it take?
That question needs the process, not the product. So the process is what we collect.
Gets in
A real thing people can open
A live URL, an app store listing, or a public repository. If a reader cannot reach it, it is not a record.
AI did material work on it
Not a spell-check, not a README. The model had to be load-bearing in the building of it.
The builder will talk about the process
Enough to fill a build log: what they used, roughly what it cost, and where it went wrong.
The submitter owns it
Your own project, or one you have the builder’s permission to submit. No scraping other people’s launches.
Does not get in
A screenshot you cannot click through to
The single most common submission.
A thin wrapper around one chat API
A blank text box with a system prompt is not a project.
An aggregator of other AI projects
This is what we are. We are not going to list our own competitors.
Anything the submitter does not own
Including “inspired by” clones of someone else’s launch.
Generated content farms
Sites whose pages exist to rank, not to be read.
Crypto, gambling, and grey-market tooling
Out of scope regardless of how well built it is.
From submission to record
Five steps01
You send the link
Through the form. A live URL or a repo, and a sentence on what AI did.
02
We check it opens
If it does not load, or it is a parked page, the process stops here. This filters out most of it.
03
We dig into the public record
The repository, the commit history, the changelog, launch posts, whatever the builder has already said. That is where the build log comes from — not from our guesses.
04
We write the record
Every claim is tied to a source, and the record shows which ones. Where a number cannot be established, it says so instead of estimating.
05
It goes in the archive
Credited to the builder, linked to the project. If the builder asks us to change or remove it, we do, without argument.
Editorial rules
Nothing is sponsored
No tool pays to appear in the stack index, and no project pays to be filed. If that ever changes, the change will be announced on this page before it takes effect, not after.
Everything is sourced
Every fact in a record is drawn from a public source — the project’s own repository, changelog, documentation or press coverage. Details can still be incomplete or imprecise. Where a figure could not be established, the record leaves it out rather than estimating it.
Builders keep their rights
You own your project and your words. Submitting grants us permission to publish the record and to quote you in it — nothing else. Ask and we will remove a record.
Corrections
If a record is wrong — a number, a date, a credit, or the whole framing — say so and it gets fixed, with the change noted if it was substantive. Nothing here is worth defending at the cost of being accurate.
Where the facts come from
Found something wrong, or want a record removed?
Everything here is compiled from public sources, so it is not necessarily 100% accurate. Treat dates, figures and quotes as a starting point and check the linked sources before relying on them.
Errata
Found an error, an out-of-date figure or a wrong credit? Send the record number and, if you have it, the source that contradicts us. Corrections are applied to the record itself and noted here.
Submitting
The form asks for a link, a sentence on what AI did, and how to reach you. Everything else comes from researching the public record — you will not be asked to write the record yourself.