Launching today
3AM Super Intelligence
Production incidents: investigated, proven, fixed
82 followers
Production incidents: investigated, proven, fixed
82 followers
Meet 3AM, the AI on-call engineer for systems you can’t rewrite. It finds the root cause, rehearses the fix, and resolves the incident with one tap in Slack, all from inside your network.
Free Options
Launch Team / Built With
Dropbox SignSimple, secure eSignatures for the way your team works.
Promoted
We're launching 3AM today: an AI on-call engineer for the systems you can't rewrite. Three things shaped it:
Engineers needed proof, not a guess. Nobody runs a fix on production at 3am because a chatbot sounded confident. So 3AM names a root cause only once a live check against your systems proves it. When it can't prove one, it says so and hands over what it found.
Regulated teams needed it to stay inside. Banks, insurers and telcos can't send production data to someone else's cloud, and their regulators ask where every byte goes. So 3AM runs inside your network with no inbound ports, works fully air-gapped too, and the only thing that ever reaches us is a signed licence check. Literally.
Everyone needed to stay in charge. Nobody wants an AI changing production on its own in the middle of the night. So every fix is rehearsed with a dry run first, then waits for a person's approval in Slack, Symphony, PagerDuty, xMatters or email (you name it). 3AM starts in shadow mode: it watches, proposes fixes and changes nothing. Giving it more autonomy takes two approvers. And every step goes on a hash-chained audit log your auditors can verify.
Proof is the point. We test 3AM by breaking a real open-source core banking system, 244,000 lines of ten-year-old Java, on purpose, in 67 different ways. Every number we publish comes from those runs, and zero harmful actions is a release gate, not a target.
If you or your team is fed up of prod being down at night (or even during the day), 3AM is for you!
🎁 As a bonus, the first 90 days are on us! Try it, tell us where it falls short, and we'll fix it.
Note that 3AM is AI, and in its open alpha stage. We'd love to personally reach out to everyone to collect as much feedback as possible, and improvise further.
How do I know it will not create more bugs than it solves ?
@bengeekly Great question and here's my honest answer: any tool that acts on prod carries risk, so we designed 3AM around never acting on a guess! Every service starts in shadow mode, where 3AM only proposes fixes and doesn't touch anything. It is also capable of building indirect relationships between various services that you have deployed. Plus, it only acts after its proven the evidence and a human has approved it. Later it will also recheck whether the fix worked. Happy to show you a live run!
hehe was there! know that pain and I'm sure many founders will feel the same. Wish you all the best here team!
@german_merlo1 Haha, then you know exactly why we're building this! Thanks a lot for the support. Really appreciate it!
"proven, not guessed" is the bit I'd want. we had prod freeze twice after a deploy and every dashboard said the app was hung, but the real cause was a schema change at startup quietly waiting on a lock behind one long running query. would 3AM's live check find that kind of thing (a waiter, not an error)? and what does rehearsing the fix look like when the fix is killing someone else's query?
3AM can inspect live database sessions and lock waiters to help identify what's actually blocking things, rather than just reporting that the app is hung.
Killing a blocking query is a different story, though. It can roll back a transaction, so we'd want to show the evidence and get approval before taking action.
@anujbarve makes sense, showing the evidence and asking first is the right call. what fixed it for us in the end was setting a short lock timeout on startup migrations, so the deploy fails fast and loudly instead of hanging the whole app. would be a nice thing for 3AM to suggest as the prevention step after it finds the blocker