Raw LLM Responses

I don't see why this is scary. I don't see why this is surprising. I don't know exactly what code was added, but it sounds very simple. If I look at a large piece of work I recently did, then I'd love to see an A.I do anything but cause a mess. Let's see, how about this one: \- Add an outbox table in its own schema to the database. \- When X happens, serialize the record to Protobuf and write to the outbox table. Wrap it in a transaction. \- Create a standalone event publisher that polls the outbox table and publish the events to Kafka. \- Metrics, monitors, etc. Now even if an A.I does all of this, and manages to do so in a reliable, resilient way, no-one would know what the hell is going on in the code. And then when something goes wrong you'll have deer-in-headlights "engineers" who are panicking and can't fix it. They won't be able to reliably verify that everything is working. They won't know if there is data integrity. They won't be able to reason about design changes, tech debt, testing, etc. Ultimately, I would love an A.I to do the small, trivial things. But I am not holding my breath for A.I to completely make me redundant by doing something described above. If that ever happens then I'll have so much domain expertise that the company would need me. For those wondering, the reason for the above outbox pattern is the dual-write problem. Only experienced engineers will understand what that is about. Not some vibe coder who hasn't built serious software. TLDR; build serious software and stop worrying.