Software engineering as I studied it at university is largely solved. Changing button colours, writing backend functions that read and write to databases, building basic websites like this one: these are all things a modern LLM produces with exquisite detail, tests and flair.
They’re also the things that, if you studied computer science (or, in my case, computer engineering), you reasonably assumed you’d spend a good part of your career doing.
Now, you could say I’m naive and that technology is always changing. But not usually this fast. Those tasks are being automated right now.
“But someone still has to do the architecture”
But wait: someone smarter than you said software engineers will always have a place doing architecture work, checking the AI’s output or cleaning up its slop code. And they’re probably right. For 18 months, max.
The frontier is moving so fast, and LLMs are getting so good at cybersecurity, architecture, refactoring and design, that the things they were bad at four months ago, when you last tested the free version of ChatGPT, are probably solved problems now.
- GPT-2 · 2019
- GPT-3 · 2020
- GPT-4 · 2023
- Claude 3.7 · 2025
- GPT-5 · 2025
- GPT-5.2 · Dec 2025
- Opus 4.6 · Feb 2026
- Mythos · Apr 2026
Measured in how long the task takes a human expert. The trend has been doubling every few months.
Source: METR time horizons
The things AI was bad at four months ago are probably solved problems now.
And yes, I totally agree that diffusing this tech into human-run organisations and processes is going to take time, and lots of people will resist handing large parts of their job to robots. But human laziness, and a competitive market that rewards productivity, will ultimately force companies’ hands, in my opinion, and make this a reality.
That’s not even to mention the strides AI is making in mathematics, drug discovery, materials science, CAD and every other industry where something is produced on a computer. Which is nearly every industry. It’s coming for all of them, and software engineering was just the first.
So what do we do?
I’m totally sympathetic to people being scared of AI and losing their jobs. By university-me’s definition, I’ve lost mine. I’ve moved on to higher-leverage work, using AI to do what five years ago would have taken a team of 20 and millions of dollars. But I understand not everyone is so keen to be kept on their toes.
This is why discussions about UBI and human meaning in an automated world are so important. If you’re stressed about being bored or unemployable, comparative advantage says there will always be something for humans to do, because the machines will be better occupied elsewhere with whatever compute is available.
But I’d go further: humans are meant for more than work. As a workaholic myself, I sympathise with defining yourself by your work in a world that rewards it. But humans are capable of so much art, love, exploration and connection that to spend our years fighting for the right to sit behind a computer… well.
I’d rather learn to paint.
The upside
The final point I’d like to make is that AI isn’t just going to cause job losses. It’s going to make so many advances in every field that the whole world will change. All of these are on the table for an intelligence that thinks orders of magnitude faster than we do, and can be scaled without the 20 years of biological bootloading the average human goes through (adolescence):
- Curing ageing, and most diseases
- Near-unlimited terrestrial energy
- Travel around the solar system
- Colonies on Mars
These are worthwhile goals to give up your office job for, and a brave new world worth embracing. Let’s see what happens, and do our best to make it a reality.