A software engineer's post climbed to the top of Hacker News and gathered 687 comments. The title was direct: "LLMs are eroding my career and I don't know what to do." What the author described wasn't a layoff notice. He still has a job. The paychecks still arrive. And yet there's a sense that something has changed — a sense that 729 people read and then upvoted.

It's worth looking a little closer at where that recognition comes from. If this had been the vague fear that "AI is taking our jobs," the comment thread would have looked different. There was reassurance, there were counterarguments, but far more common were firsthand accounts of "I know exactly what this feels like." What the post touched wasn't some distant future scenario. It was a concrete change people feel in front of the tools they use every single day, and that is what generated 687 responses.

What changes when you become the person reviewing the code

Using an LLM makes writing code faster. That much isn't up for debate. A large share of engineering work now arrives as an LLM-generated first draft, and the human shifts into the role of reviewing and revising it. On the surface, this is efficiency. You can get through more work in the same amount of time.

But the point where the author started to feel uneasy lay somewhere else. Before, writing code from scratch meant working through, by hand, why it behaved the way it did and where it might break. Deciding line by line was tedious and slow — yet that tedium was the foundation that made it possible to handle more complex systems later. As the LLM supplies the first draft, that path gets shorter, and the practice of designing and constructing the logic yourself diminishes.

What happens after six months, or a year? When a situation arrives where you have to write code from scratch without an LLM, can you still do it at the old pace? The author wrote that he wasn't sure. He didn't use the phrase "the muscle has gone stiff," but the feeling was close to it.

This phenomenon isn't unique to software engineering. The planner who has started handing planning documents to an AI, the practitioner who has begun drafting reports with an LLM — they end up facing a similar question. Do I still hold the logic of this work inside my own head? If someone asked why this sentence is here, or why this structure turned out the way it did, could I point to the reason?

Layered on top of this is a signal from the market. If one AI-assisted engineer can handle what used to take three people, companies gain an incentive to resize their headcount. From the second half of 2024 onward, a string of technology companies — Google, Meta, Salesforce among them — scaled back hiring or trimmed team sizes, and a fair number of these moves were restructurings tied to AI adoption. That's the backdrop suggesting the author's anxiety may not be purely a matter of personal sensitivity.

The counterargument: that LLMs actually raise a person's value

There is a case running in the opposite direction, and it can't be brushed aside lightly.

As several engineers pointed out in the comments, LLM-generated code can't be pushed to production unless a capable person reviews it. Making architectural decisions, responding to shifting requirements, spotting security vulnerabilities and edge cases — these remain the human's job. From this vantage point, the LLM relieves you of repetitive work so you can concentrate on harder problems. Not a crisis, but a reshuffling of roles.

History offers similar arcs. When Excel arrived, people predicted accountants would vanish; in reality, demand for handling more complex financial analysis grew. When Photoshop arrived, there were fears that graphic designers would dwindle, yet the design industry itself expanded. There's a recurring pattern in which technology automates some roles while simultaneously creating demand for higher-level ones.

But for this counterargument to hold, one condition has to be met. It's true for the person who already has enough capability to "concentrate on the harder problems." For the person still building that capability — or the person who, in the act of handing repetitive work to AI, also handed over the practice of thinking for themselves — the outcome is different. The reason that 729-point confession resonated so uncomfortably is that it carried the second of those feelings.

What changes for Korea's solo operators and middle managers

Move this discussion into the perspective of a Korean practitioner and the question narrows to one. As the share of work I'm handing to AI grows, is the value I add visible from the outside?

Middle managers driving AI adoption all run into a common situation: throughput rises, but the roles of team members blur. When the workload drops while headcount stays the same, pressure builds to explain to higher-ups what each person is actually doing. "The AI did it" doesn't relieve that pressure. What has to be visible is which direction you steered that AI, where you intervened, and what you added to the result.

From the standpoint of running a team, another worry arises. If team members have sped up by using AI, where should that freed-up time go to become a path toward more growth? Does processing a larger volume of the same work build a team member's capability — or is using that time to deeply understand the logic and background of the work done so far the more sustainable investment? A manager's answer to this question steers the whole team in a different direction one or two years out.

For a solo operator, a different kind of pressure appears. If you used AI to speed up the work, how did you fill the time you saved? If you moved toward processing more of the same kind of work, you may find yourself drawn into price competition or pushed for ever-shorter turnaround times. From the client's side, when a freelancer who uses AI and one who doesn't can't be told apart by the deliverable, the basis for pricing changes.

Hiring is shifting too. People who have spent years in recruiting tend to mention the same thing: the focus has moved away from which tools you can use toward how a person approaches a problem and what context went into the work. Applicants who attach a separate piece of writing — "why I chose this direction in this project" — rather than just a portfolio are getting a different reception in interviews.

I don't think this is merely a matter of self-promotion skill. Being able to write that account is evidence that you still hold the logic of your own work inside your head. If, even while letting AI handle the first draft, you can say where and how that draft fell short and in which direction you revised it, that is a signal that your capability is still alive.

One check helps. Of the work you handed to AI over the past month, what percentage can you explain in terms of how it differed from doing it yourself? If you can't explain it, it means the value you added in that work isn't visible from the outside. Whether it's a recruiter, a client, or a team lead, they all ask the same question: what does this person add on top of the AI?

The confession that 729 people recognized on HN was a shared feeling among those who, in riding the speed of AI, have gradually cut back on the act of thinking for themselves. The faster AI makes the work go, the more visible the person becomes who can point to where they intervened and what they changed.