Prelint is Live on Product Hunt, Need Your Support! 🙌

by

Prelint is live on Product Hunt today! 🚀

We'd really appreciate your support. If you have a moment, please check it out and share your thoughts.


🔗

Every upvote, comment, and share means a lot.


Thank you for your support! ❤️

111 views

Add a comment

Replies

Best

the bench ranking and live org chart are solid proof, still want to see it run on something client-facing before I'd trust it there..

 thanks for the comment! Honestly, we built it mostly for ourselves originally because other tools didn't catch silent drift like code that works but completely ignores design decisions around architecture or code design patterns.

Then we shared it with first clients and it worked surprisingly well (where it's enabled it's quoted in about 40% of all fixes on Pull Requests) so a great way to try it out is to start running it on some first projects and you should see it being quite helpful pretty much immediately - it's plug and play.

The cool thing is, it surfaces all the implicit decisions that coding agents make on your behalf so you should see an "aha!" moment with the first decision reviews it posts.

   That “client-facing before I’d trust it” is the nicest boundary. We see a parallel kind of drift: code can stay aligned with architecture while the product decision drifts away from the customer reason that creates it.

The review is much stronger when it carries both the change and the evidence/intent it is meant to preserve—and what would count as success or failure. Curious whether Prelint sees a path to carrying that upstream product context too, or is its boundary intentionally the code/design layer?

You're 100% right Jorge. One of the things I love about Prelint is that it's very accessible to non-technical people too and - given the velocity most AI-pilled teams move with now - it makes the decisions getting baked into the product much easier to understand.

So it's really helpful at not only answering the question "is our architecture/product drifting" but also "how is it affecting our customers" and most importantly getting this message across to the relevant people, in a way that they understand.

Somewhere downstream we have voice interfaces for non technical stakeholders to interact with our "decision ledger" planned in our backlog. The people interacting with the business and customers often don't even have GitHub access and I see it as our job to give them the right tools,

Nice boundary. Voice IO feels like a nice add-on. The people closest to the customer often don’t have access to the repo or PR as well. The shared question I come back to is how to keep that revenue and customer connected to the decision without forcing it into an engineering workflow.

An in, if the decision ledger can carry the original customer reason and the outcome back to them, it could close the loop in both directions. Most teams focus on the build so much that they loose sight of the customer, where product drift really happens most on early decisions, before even a line of a requirement is written. Let's compare notes one day!

Following the thread with Jorge. The distinction that's interesting to me is who's positioned to catch drift first. Prelint catches it at the PR, which is already further upstream than most teams check. But the earliest tell is usually behavioral, not architectural: an agent quietly reinterprets a spec in a way that's locally sensible and globally wrong, days before it ever becomes a diff.

Curious whether the decision ledger idea eventually surfaces pattern drift not just "this PR contradicts that ADR," but "this is the third time an agent has independently made this same reinterpretation," which is usually the signal that the spec itself is ambiguous, not the code.

Wrote up more of this; the "agents aren't making style mistakes anymore, they're making architecture decisions" framing in a thread this morning if useful:

Congrats on the launch, this is a real problem and good to see it shipped plug-and-play rather than as another dashboard nobody opens.