Blog

Notes on on-call, reliability, and AI

From the team building Ketl: thinking out loud about incidents, language models, and what it takes to make on-call less painful.

ReliabilityLatest

The anatomy of a 3am page

Most of an incident is not the fix. It is the blind digging that comes first. Here is where the minutes actually go, and why a fast first read changes the shape of the whole night.

Marisol ReyesJuly 14, 20266 min read

Engineering

Reading logs is a language problem

Telemetry is messy, unstructured, and different at every company. That is exactly the kind of problem modern language models are good at, and exactly why the model is the engine of Ketl, not a feature on the side.

Aaron FeldJune 24, 20267 min

Models

Why we route across models

Betting a pager on a single model is a reliability risk of its own. Ketl keeps a registry of frontier and open models and routes each step of a diagnosis to the one that clears the bar for the least cost.

Devin ChoJune 3, 20266 min

No more 3am guesswork.

Connect your alerts and watch Ketl read the first draft of every incident. Free to start, no card required.

Start freeSee how it works