Aamer Mihaysi

I read the post that's making the rounds today — the one where a software engineer says LLMs are eroding their career. It's not a hot take. It's a confession, and it hit me harder than I expected.

The author isn't complaining about code quality or job loss. They're describing something more insidious: the slow hollowing out of the work they used to love. The feeling that the problems worth solving are getting eaten by autocomplete, and what's left is glue code and prompt-tweaking. They're not wrong.

I've felt it too. There's a specific kind of dread that creeps in when you realize the hardest debugging session you'll have this week isn't a race condition — it's figuring out why the agent refused to call a tool on the third retry. The craft is shifting under our feet, and pretending it isn't doesn't help anyone.

Here's what I keep coming back to: the work isn't disappearing, but the shape of it is changing fast. The engineer who used to spend three days writing a parser now spends those three days building a validation layer around an LLM's output. Different muscle, same problem — reliability. The difference is the old parser was deterministic. The new one hallucinates, and you have to design around that.

The post resonated because it names something most of us won't say out loud: we're not sure what "good" looks like anymore. A clean PR used to mean readable, tested, efficient code. Now it means the agent didn't introduce a subtle logic error in the third function, and the tests pass, and you think the output is right but you're not 100% sure because the model's reasoning is a black box.

I don't have a tidy answer. But I think the way through is to stop treating LLMs as junior engineers and start treating them as what they are: stochastic text generators that sometimes write correct code. That means we own the verification layer. We own the architecture. We own the decision about when to let the model run and when to hardcode the logic.

The career isn't eroding. It's relocating. The engineers who figure out where it's going — and get there first — are going to be fine. The ones who wait for someone to draw the map? Those are the ones I'm worried about.

https://human-in-the-loop.bearblog.dev/llms-are-eroding-my-software-engineering-career-and-i-dont-know-what-to-do/