Ask HN: SWEs how do you future-proof your career in light of LLMs?

> The more I speak with fellow engineers, the more I hear that some of them are either using AI to help them code, or feed entire projects to AI and let the AI code, while they do code review and adjustments.

I don't see this trend. It just sounds like a weird thing to say, it fundamentally misunderstands what the job is

From my experience, software engineering is a lot more human than how it gets portrayed in the media. You learn the business you're working with, who the stakeholders are, who needs what, how to communicate your changes and to whom. You're solving problems for other people. In order to do that, you have to understand what their needs are

Maybe this reflects my own experience at a big company where there's more back and forth to deal with. It's not glamorous or technically impressive, but no company is perfect

If what companies really want is just some cheap way to shovel code, LLMs are more expensive and less effective than the other well known way of cheaping out

There's no reality in the next twenty years where a non-technical individual is going to converse with a persistent agentic AI to produce a typical SaaS product and maintain it over a period of many years. I think we'll see stabs at this space, and I think we'll see some companies try to replace engineering teams wholesale, and these attempts will almost universally feel kinda sorta ok for the first N weeks, then nosedive and inevitably result in the re-hiring of humans (and probably the failure of many projects and businesses as well).

Klarna said they stopped hiring a year ago because AI solved all their problems [1]. That's why they have 55 job openings right now, obviously [2] (including quite a few listed as "Contractor"; the utterly classic "we fucked up our staffing"). This kind of disconnect isn't even surprising; its exactly what I predict. Business leadership nowadays is so far disconnected from the reality of what's happening day-to-day in their businesses that they say things like this with total authenticity, they get a bunch of nods, and things just keep going the way they've been going. Benioff at Salesforce said basically the same thing. These are, put simply, people who have such low legibility on the mechanism of how their business makes money that they believe they understand how it can be replaced; and they're surrounded by men who nod and say "oh yes mark, yes of course we'll stop hiring engineers" then somehow conveniently that message never makes it to HR because those yes-men who surround him are the real people who run the business.

AI cannot replace people; AI augments people. If you say you've stopped hiring thanks to AI, what you're really saying is that your growth has stalled. The AI might grant your existing workforce an N% boon to productivity, but that's a one-time boon barring any major model breakthroughs (don't count on it). If you want to unlock more growth, you'll need to hire more people, but what you're stating is that you don't think more growth is in the cards for your business.

That's what these leaders are saying, at the end of the day; and its a reflection of the macroeconomic climate, not of the impacts of AI. These are dead businesses. They'll lumber along for decades, but their growth is gone.

[1] https://finance.yahoo.com/news/klarna-stopped-hiring-ago-rep...

[2] https://klarnagroup.teamtailor.com/#jobs

It’s simple, as soon as it can write code faster than you can read it you can’t trust it anymore. Because you can’t verify it: Trusting trust.

If it keeps you busy reading code, if it keeps all of us busy consuming its output. That is how so will conquer us. Drowning us in personal highly engaging content.

Stop using llms, if they can solve your problem than your problem is not worth solving.

LLMs are great for simple, common, boring things.

As soon as you have something less common, it will give you widely incorrect garbage that does not make any sense. Even worse, it APPEARS correct, but it won’t work or will do something else completely.

And I use 4o and o1 every day. Mostly for boilerplate and boring stuff.

I have colleagues that submit ChatGPT generated code and I’ll immediately recognize it because it is just very bad. The colleague would have tested it, so the code does work, but it is always bad, weird or otherwise unusual. Functional, but not nice.

ChatGPT can give very complicated solutions to things that can be solved with a one-liner.

This is a real thing. LLMs are tools, not humans. They truly do bring interesting, bigger problems.

Have people seen some of the recent software being churned out? Hint, it's not all GenAI bubblespit. A lot of it is killer, legitimately good stuff.