Post

both software engineers’ and mathematicians’ jobs have been mostly automated, but for the latter it feels catastrophic but not for engineers why? imo it’s because mathematicians relied on attribution & honor, whereas for software it’s easily reduced to the net value produced

4 reposts95 likes

neither job is automated (software and understanding aren't cheap), but there's progress generating what was historically a key part of the work product (code and proofs) software cos get that they have to draw the rest of the owl (qa, high-level design, etc.). mathematicians want the same thing

i think the state of things is also different between the two fields--the proofs are apparently basically unreadable, like how llm code was once much worse. 'astra solved math' feels kinda like 'sonnet 3.7 solved coding'

yeah, i just see a lot of engineers embracing change that are in the best positions of their careers yet

I think a big part of it is that “solving a millennium prize problem” shows how large the gap is between humans and AI in math in an easily understood way, but there has been no such event for software engineers. some engineers still insist AI can’t code because of this.

I'd say it's been bad for software engineers too; lack of junior role hiring, excessive firing and reduced demand of labor in favor of senior dev running herds of agents

IME, I think it's really the middle that's suffering right now. Junior engineer (interns, new grad) pipelines are still active. Experienced people are highly in demand. Someone with 2-4 years of experience is in a tough spot - a new grad is cheaper and senior/staff gets you better bang/buck.

i’ve heard mixed messages on the junior part. i think you’re probably right, but some insist that older people tend not to understand AI like younger people do. i guess we’ll see how it pans out long term

Younger folks also have had AI/ML fundamentals as a core part of the curriculum for some years now. I sure as hell didn't do any ML when I got my CS degree 17 years ago. They're far better poised to take advantage of the innovative gaps around LLMs, over simply automating code production.

Math and coding hasn't been automated, even by models as advanced as Fable I still have to deal with terrible code reuse and catching it's errors. Claude Code itself has bugs in it

Automation may not mean you have less work. Automation is measured by increase in output. Are you seeing more of that, or not? Honest question, for me its a gamechanger, but your POv may differ, and I’d like to understand why.

Blind output or productive output? I would have hoped we learned decades ago not to measure software quality by quantity, "Measuring programming progress by lines of code is like measuring aircraft building progress by weight" - Bill Gates. Do we need a book called The Mythical Bot Month?

Productive output. I have running software here build in a month by 1 or 2 guys that would have taken years to build with a 10 person crew. I can’t claim 10x increase in all cases, but it happens, and LOC is not the measure here.

It is simply not true that mathematicians’ or software engineers’ jobs have been mostly automated

Also lot of software engineering is about automating things. That is why wages are higher. For same reason I think software engineering will be one of the last professions, because lot of people will move to automate physical world with robots, when there isn't work left in the digital realm.

I don't think that's the case at all. Agree it might feel a bit more like a bomb shell now but that's because there has been great leaps recently so it feels like AI is "now" able to do maths, LLM engineering capability has grown gradually by virtue of being used case zero

Parts of maths will be massively spread up, we'll have AI proof factories spitting out results all over the place. Fine. But the pursuit of maths is about the journey. The whole point is that humans are not just along for the ride like a lot of software engineering is. It's not purely outcome lead

yeah, there’s an adjustment period, but at this point i think it’s starting to look less and less catestrophic