GenClassA race hunt on your app. US$500, or your money back.
We run your web app in a real browser while GenClass delivers its responses in other orders, and look for the moments where the page ends up wrong: an older response applied over a newer one, a form sent twice, an edit that is lost. You get every race we could reproduce, each with the steps, the order that triggers it and a fix. If we reproduce nothing, we refund you in full.
For every race we reproduce, six things.
The steps
What it isWhat to click, in order, to see it on your own machine.
What it is forYou can check it by hand before you believe us.
The order
What it isWhich response arrived before which, and the seed that replays that order.
What it is forIt reproduces every time, not one time in fifty.
A recording
What it isThe page going wrong, on video, next to the same steps in the usual order.
What it is forAnyone on your team can see what a user would see.
The cause
What it isThe lines in your code where the older response wins.
What it is forThe fix is a small change, not a search.
A fix
What it isA pull request if you give us access to the repository, a patch if you would rather not.
What it is forYou review it like any other change.
A regression test
What it isA Playwright test that replays the order. It fails before the fix and passes after it.
What it is forThe race stays fixed.
Five working days, and a person checks every finding.
You pay and point us at the app
What happensOne payment by card. No subscription and no account.
What we need from youA staging address with a test login, or a repository that starts with one command.
We hunt
What happensAn AI agent drives your app in Chromium while GenClass holds responses and releases them in other orders. Where the page ends differently, it records the order.
What we need from youNothing. We use test data only and never touch production.
A person confirms
What happensSomeone on our side reproduces each finding by hand. What cannot be reproduced is left out of the report.
What we need from youNothing.
You get the report
What happensWithin five working days of the day we can sign in to the app. If we reproduced nothing, we refund the payment in full and tell you what we tried.
What we need from youAn address for the report, and a reviewer for the pull requests.
46 open-source apps hunted. 45 had a bug we could reproduce.
In October and November 2026 we hunted 46 open-source web apps and sent 45 reports. Maintainers have fixed 16 of them, among them Gitea, React Admin, Keystone, LibreChat, Langflow, AnythingLLM, Memos, Notesnook and Umami.
Not every bug was GenClass's
The numberOf the 92 bugs we confirmed, GenClass flagged 45. The rest were found while it ran, or by reading the code.
What it means for youYou are buying the hunt and its result, not a claim about one tool.
The explorer has false alarms
The numberIn these hunts it flagged a race on a page that was correct in 197 of 745 reordered runs.
What it means for youThat is why a person confirms every finding before you see it.
We ask first now
The numberThose 45 reports were sent without asking, and some maintainers objected (issue 173).
What it means for youA hunt happens only when you ask for one, and the report goes only to you.
Before you buy
What kind of app can you hunt?
A web app with a browser interface that talks to an API: React, Next.js, Vue, Nuxt, Svelte, Angular or plain JavaScript. It has to run somewhere we can reach, or start from its repository with one command.
Do you see our users' data?
No. We work on a staging or development copy with test data and a test login. We do not need production and will not ask for it.
What counts as "nothing reproducible"?
A finding counts when the same steps and the same order of responses make the page end up wrong every time we run them, and a person on our side has seen it by hand. If there is no such finding, you get the full payment back.
Is this a subscription?
No. One payment for one hunt on one app. The runtime, the dashboard, the MCP server and the GitHub App stay free.
Our app cannot leave our network. Now what?
Write to mehar@daybot.ca. The hunt can run inside your network on terms we agree.
Who is behind this?
GenClass is made by Daybot Solutions Inc., in Canada, and your card statement will say DAYBOT.CA. The runtime and the collector are open source. The terms of a hunt are in the terms, and what we receive is in the privacy policy.