Theo says agents triaged his outages and rolled dozens of API keys
Days after saying his vibe coded work collapsed, Theo quote-posted himself with the opposite experience.
Theo quote-posted his own post from earlier in the week and said he had the exact opposite experience. "I would never have survived the day without agents," he wrote. He says they caught a lot of problems, triaged all the outages and security issues, and rolled dozens of API keys across various platforms.
They triaged all the outages and security issues.
I would never have survived the day without agents.
The post he was replying to was his own, from two nights earlier, where he described one of those nights where all the vibe coded pieces collapse and concluded engineers are still very, very much necessary. He has not retracted that one. Both posts are on the timeline, days apart, pointing in opposite directions.
What the posts actually cover
The two claims are about different jobs. The collapse was about code that agents wrote and that then fell over. The rescue is incident response, which is triage, key rotation and chasing down security issues across platforms. Those are different failure modes and different skill demands, and nothing in either post says the same workflow produced both outcomes.
The rolling of dozens of API keys is the most concrete detail he gives. It is also the kind of task that is mostly mechanical and wide rather than deep, repeated across many dashboards and services, where an agent loop with credentials and patience is plausibly faster than a tired human doing it by hand at the end of a bad day.
What is missing
There are no receipts here. Theo does not name which agents he used, which models, what the outages were, how many keys exactly, or whether the outages were in systems the agents had previously built. He does not say how he verified that the rotations were complete or that nothing was left pointing at a dead key. The post ends with "Building with AI is pretty damn cool," which is a mood, not a measurement.
For working engineers, treat this as one person's account of one bad day rather than evidence that agents are good at incident response. The useful read is narrower. An engineer who had just watched agent-written code fall apart still reached for agents the moment the job changed from writing features to cleaning up a mess, and found that the second job went better than the first. Whether that holds beyond his own platforms is not something either post answers.
Have the good opinion, not the loud one
One email a day. Under three minutes. What working engineers found out about coding with AI.