Where do we go from here?

AI has changed the software engineering profession. Is this change permanent? If so, what does the profession look like moving forward? Are incoming grads screwed? Does experience matter at all? What should I do about my career if I'm not ready for early retirement, but am too old to make a significant career pivot?

I'll explore all of this and more below.

Is this permanent?

I think so, yeah. How could this all grind to a halt and reverse? The AI companies could go belly-up, but in that case, their existing models would be bought on the cheap by a big corp like Microsoft. The models won't disappear due to bankruptcy.

What about poisoning? Some disgruntled devs are attempting to disseminate LLM poison across the web. I think this is both wrong and futile. If the malcontents succeed in poisoning all future training attempts, the current models will become the permanent plateau. They're good enough as is, so poisoning won't give us a reversion to the good old days.

What does software engineering become?

The majority of work in this industry will be a combination of product management, micro-management of agent teams, and code reviewing.

A note about AI skepticism

Here, I'll take a moment to address AI skeptics. The skeptics claim that AI is only good for generating boilerplate code. It can't do anything novel or interesting. The only jobs it will automate are code-monkey jobs, and the interesting jobs will remain.

But, I think this is wrong. With guidance, agents can definitely generate code which has never existed before. If you think of a completely new concept, you will still build it on existing languages using variables, conditionals, and loops, or some higher level functional abstraction over those. Well, if you can write it, you can guide an agent to write it.

I've gotten Claude Code to write most of a password manager in Hare-- an obscure language that is almost certainly a tiny fraction of Claude's training data. It can do more than build trivial CRUD applications.

Are incoming grads screwed?

I don't know. Maybe. My advice to young people is to try to find work that they'll enjoy, that will contribute positively to society, and that will not be easy to automate away. Blue collar work clearly fits the bill: plumbers, builders, electricians, etc. Careers that will always require human discretion, or where human-to-human interaction is part of the point-- doctors, nurses, psychologists, etc.

If you're entrepreneurial, think about how you can provide value to people in your community. Be a value creator, and ideally, find a way to create value in the physical world.

Does experience matter at all?

Yeah. It does. The best way to guide an agent is to break a system down into clean modules with clear boundaries, so the agent can focus on specific subsystems without blowing up context or getting lost in the weeds. Without proper organization, a project will slowly become unmaintainable. This is true whether regardless of whether it's built by humans or by AI agents.

What should I do about my career?

Personally, I've avoided climbing the management ladder. I've been a team lead a few times, but that mostly involved hiring the right people and letting them do what they do best. I don't know that I want to spend the rest of my career coddling and babysitting LLMs. I don't see myself lasting very long if my job consists primarily in reviewing AI slop.

So, what's an engineer to do? To this, I don't yet have an answer. This year, I will discover whether or not I enjoy shifting primarily to product management.