Product · Argus

Finding the incident you've already solved

Most incidents are not new. Somewhere in your team's history is the same failure, already diagnosed, already fixed — if only anyone could find it at three in the morning. That memory is the idea behind Argus's similar-incident search and drafted postmortems.

You have solved this before

The cruel thing about operational failures is how often they repeat. The database fills up again, the certificate expires again, the same downstream service times out under the same load. The fix exists — someone worked it out, painfully, months ago — but it is buried in a chat thread or a half-finished document, effectively lost.

So the team pays for the same lesson twice. The knowledge was there; it was just not reachable in the moment it was needed. An operations tool that cannot help you find your own past is asking every engineer to hold the whole history in their head, which is exactly what fails at the worst hour.

The idea behind similar-incident search

Argus is being built so that when an incident opens, it can surface the past incidents that look like it — not by exact keyword match, but by resemblance, the way a colleague who has been around a while says this feels like that thing from March. The goal is to put the earlier diagnosis in front of you while it is still useful.

That is what similar-incident search is reaching for: turning a team's accumulated scar tissue into something searchable in plain English. It is a proof of concept, not a finished promise — but the idea is straightforward and worth stating clearly, because the value is obvious to anyone who has re-solved a solved problem.

The postmortem that starts itself

The other half of the idea is the postmortem. Everyone agrees they matter; almost nobody wants to write one after a long night. So the write-up gets skipped, and the lesson never enters the record that similar-incident search would later draw on. The gap feeds itself.

Argus is being built to draft the postmortem from what the incident already knows — the timeline, the actions taken, the resolution — so a human edits rather than starts from a blank page. Do that reliably, and each incident feeds the memory that helps solve the next one. That loop is the whole point, and it is what the proof of concept is trying to earn.

Questions

People also ask

    What is similar-incident search?

    The idea, in Argus, of surfacing past incidents that resemble a new one — by likeness rather than exact keyword — so an earlier diagnosis is in front of you while it is still useful. Argus is a proof of concept; this describes what it is being built to do.

    Does Argus write postmortems automatically?

    It is being built to draft a postmortem from what an incident already records — the timeline, actions, and resolution — so a person edits rather than starts from scratch. It is a proof of concept, not a live product.