Hey Product Hunt,

Take-homes used to measure how quickly someone could produce working code. That signal has thinned out. The code arrives either way now, and it usually runs.

What doesn't arrive with the code is the thinking behind it. Where they look first, what they rule out, whether they read the error or guess at it.

So Bisect puts candidates somewhere a finished repo can't hide: a real Linux VM that is already misbehaving.

How it works

Pick a challenge and send a link. Filter the library by stack and seniority. No account for them to create.
They work a system that is genuinely broken. Terminal, editor and file tree in the browser, backed by a sandboxed VM. 15 to 90 minutes, async, on their own machine in their own hours.
You review the replay. Every command, keystroke and file change is captured server-side, so it is what happened rather than what was reported. AI reads the session against the rubric and drafts a verdict with the evidence attached. You agree or disagree on the record. About a couple of minutes per candidate.

Why engineers prefer it

It is the format they already asked for. Async, unsupervised, nobody watching them type, no whiteboard, no trick question with one accepted answer. They debug something real, which is the job.

Why hiring managers prefer it
No senior engineer pulled out of their day to sit in on interviews. No demo call, no procurement cycle, nothing to deploy on anything of yours. You get a recording and a drafted verdict instead of a blank review page.

130+ challenges across Go, Python and PHP, junior through senior, and the library rotates so last month's challenge does not leak.

Seven days free, no card. Then £89.99/month for the Pro plan.

I would genuinely like to hear where this breaks for your team, especially if you hire outside Go, Python and PHP. Tell me which stack and I will put it in the queue.