In the mid-2000s, I chose not to study software engineering in college because my parents ran a staffing firm that outsourced SWE work to India. I had enjoyed the coding I'd done in high school but had convinced myself there would be no career stability in learning how to program. A few years later, I realized my mistake and had to play catch up.

I see a lot of the same thinking in the wild today. We're adding powerful new abstractions, but a) that doesn't mean we don't need at least some people who can see through those, and b) greater abstraction means we can take on bigger challenges, and there's no shortage of those.

We're still a very long ways away from LLMs/AI replacing SWEs. I don't even think we'll see them replace SWE in my lifetime, at least not with the way LLMs work currently.

LLMs are great at producing single functions, maybe even a whole class. But they really are terrible architects and can't produce something that requires multiple systems to interact. Their simple method of "predict the next token" is bad at planning and is mainly incapable of producing complex systems.

Until AI gets to a point where you can just say "Create an MMORPG with these features [....]", there will be a need for SWEs. Until then, it's just "Create a function that adds all the stat bonuses from the player's gear" and hope that it can understand your other code that defines the player's gear.

I'm working on an agent-based AI coding tool[1] that is trying to push the limits on the size/complexity of tasks that can be automated by LLMs, so I think about this often. I'm also using the full gamut of AI tools for development (I have 5 subscriptions and counting).

My opinion is that skilled engineers won't be replaced for a long, long time, if ever. While AI codegen will keep getting better and better, getting the best result out of it fundamentally requires knowing what to ask for—and beyond that, understanding exactly what is generated. This is because, depending on a particular prompt, there are hundreds/thousands/millions/billions of viable paths for fulfilling it. There are no "correct answers" in a large software project. Rather there is an endlessly branching web of micro-decisions with associated tradeoffs.

I think the job of engineers will gradually shift from writing code directly to navigating this web of tradeoffs.

1 - https://plandex.ai

> LLMs and Agents are not automating SWE, but they are changing it.

Until an LLM can translate opaque product requirements into a working feature, including covering the nuances and edge cases of the product, I am not worried. I have written software professionally for 12 years and am convinced people skills are what make or break a successful developer. Kindness, empathy and generally being an easy person to work with are far more important skills (IMO) than knowing how to write software.

I consider myself midcareer and do not plan on making a change anytime soon if ever. I believe as long as I need a career (probably much longer) there will be a need for people who can write software.

There may be a small bit of truth in each claim, but they are mostly false.

> LLMs and Agents are not automating SWE, but they are changing it.

LLMs are maybe a slightly larger threat to SWE careers than the ability to copy/paste code from stack overflow.

> Interviews have become nightmares

Interviews if anything have gotten slightly better. Companies stopped asking riddles like, “how many windows are in Seattle”.

> A lot of SWE work is objectively make-work, and a lot of startups are building fluff or things that are net negative to society.

This has been the case for a long long time.

> Beyond this, the supply for SWEs vs the demand is clearly imbalanced on the supply side.

Tech salaries say otherwise. They don’t pay 3-4x the average household income out of kindness.

> Is SWE a sustainable career anymore?

Definitely, though it depends on your definition of sustainable.

If you want a remote position, 30 hour work week, and a $200k+ salary maybe not.

But lots of people will continue to earn a good amount in an overall cushy job as SWEs for the foreseeable future.

Dairypharmer mentioned it but in the late 90s-early 2000s, you could find a lot of folks on Slashdot saying that there was no point majoring in computer science or learning to program because all those jobs were going to be outsourced to India soon anyway.

I'm probably more negative on this than average since I'm already x-tech and long term unemployed (middle aged) but my take is it was never a sustainable career for _most_ people. A minority get lucky, or are otherwise able to signal the right qualities to retain steady employment.

Even if you do manage to make it as an IC all the way into your forties, you're going to run into the pervasive age discrimination, which is only getting worse. And that is before AI factors and general capitalist job-destruction effects are even considered.

However if you enter the field and treat it as a money pump for a decade or two that will inevitably fizzle out, and you have a plan B career, it may not be a bad way to start. Just have that plan B ready (I did not).

My personal opinion is most software will be easy to write by a junior programmer. There will still be high paid roles, but this will be a low paid profession. Web development, UI, etc will be heavily automated with LLMs. You will guide the model in building the software. It's already this way using cursor.sh, most people just havent realized it yet.