Semfora
Semfora - AI governance in a way that linters never could.
22 followers
Semfora - AI governance in a way that linters never could.
22 followers
AI made writing code cheap. Understanding it got expensive. Semfora indexes every merge into a deterministic map of your codebase across 30 languages. Same commit, same answer, every time. Dashboards, chat that cites the code it is talking about, merge analysis, and a pull-request gate your team configures. Next: state an architectural rule once and have it enforced on every merge. Prototype today. Free tier. Connect one repo and try to catch it being wrong.
Connected one repo and asked about a sketchy module from six months ago. It pulled up the exact function and the merge that introduced it, which was honestly more than I expected. The cited answers in chat feel like a real upgrade over grepping my own history.
@lylelovett so glad to hear that, and I am sorry about the terrible chat UI lol the website took backseat to engine function.
@lylelovett Should be rolling out now, but an improved repo dashboard UI, including the automatic ontology generation. I am curious if there are any free rules you didn't expect there!
A lot of the first pass at generation is simple stuff. Services passing data down, not up. And that kind of thing.
But, when you get to the near misses. It forces you to ask questions about your codebase that will help keep AI tame for a long time.
Internally, we have seen a drastic reduction in AI mishaps in our engine, as the axioms act as gates, and documentation.
For instance, you can replicate all of schema.org in our axioms, and then use them to do "axiom driven development" which will fail until they are all true.
That has been a very fun experiment.
But, I am glad you like it! It's early days, and we have a Bunch more to build!
finally connected a repo and the chat citing the exact lines it pulled from actually convinced me, my team stops debating what that legacy module actually does