The idea of "self-driving SAAS" sounds great until you actually have to manage it. It's pretty interesting how they're using Datadog monitors to trigger the whole thing, but I'm skeptical about the scope. They've basically narrowed the entire use case down to just catching dead feature flags and fixing failing background tasks.
>The agent runs our cleanup skill and one-shots it nearly every timeThat works because those are low-context, repetitive tasks. If you tried giving it a real logic error without enough context, it'd probably just hallucinate a fix and break something else. I can see it working for cleaning up old code, but the moment it touches anything more complex than a dead flag, we're going to be spending more time auditing its PRs than actually coding. It feels like
manual oversight is still going to be the bottleneck here.
more here:
https://linear.app/now/linear-agent-bug-fix