Articles & threads
Josh Rosen rounds up the first two weeks of JevOps — decision-model calls showing up in telemetry, incident response, SRE agents, CI, and deploys — and argues DevOps is packed with exactly the small semantic judgments Jev is built for.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
The examples are early and mostly small, but the pattern is visible: instead of asking a model to reason through a whole incident, operational systems can afford thousands of narrow calls like "is this log worth investigating?" as things run. Figures are the projects', not ours — Cribl saw about 92% agreement with its judge committee at roughly 1% of the cost; SREGym's Jev-assisted agent went from 20 to 24 passes out of 50.
Start by listing repetitive checks your team does when systems run. For example, note down questions like whether an error log needs human review or if a server alert is urgent enough to wake someone up. This gives you clear rules to test before writing code.
A developer can then adapt this concept by building small model checks into your server monitoring tools. Keep in mind that this article only shares early ideas and community experiments, not ready-to-run software, so you cannot rely on it to safely resolve system emergencies alone.