I teach computer science at Montana State University. I am the father of three sons who all know I am a computer programmer and one of whom, at least, has expressed interest in the field. I love computer programming and try to communicate that love to my sons, the students in my classes and anyone else who will listen.
A question I am increasingly getting from relatives, friends and students is:
Given AI, should I still consider becoming a computer programmer?
My response to this is: “Yes, and…”
I’ll say this much: people don’t have to work for a big, publicly traded corporation. There are still smaller software houses out there where the executives aren’t little more than the shareholders’ fluffer and trust devs to know how to do their jobs, though you may need to look outside of mainstream applications. Whether they have the collective capacity to absorb everyone who wants to be a professional programmer, I don’t know. But in a world of slop, being able to provide even somewhat reliable software may be a gap in the market that could be exploited and allow for that capacity to grow.
100% My firm caters to a very niche segment and that segment requires highly reliable and optimized operations. We are beyond booked. We literally have three environment upgrades and a go live tomorrow (simultaneously) for three different clients. We clean after the vibe coders and get paid very well for our services.
My advice remains the same. There just isn’t a great future in software dev. Yes, to this day we still value custome tailored clothes and hand made, expertly crafted shoes. That doesn’t change the fact that almost all of our clothes and shoes are low quality, quasi- disposable, and 100% machine made.
I’m not sure that analogy works. The machines used for making clothes are reliable and produce repeatable results that are good enough. I recently had to throw away a 15 year old T-shirt because it was getting a bit too ratty, but it still technically functioned as a T-shirt. Also, mass produced clothing in standardised sizes didn’t actually replace the bulk of tailor-mode clothing, that was always something for the rich, but that’s getting too deep into it.
In comparison, LLM-based code generators are inherently unreliable and by their very nature incapable of ever becoming able of producing good enough results, at least with the current dominant paradigm. Many execs may not feel that way, but that’s very much a FAFO situation, because unlike clothing, where poor quality may cause it to degrade faster, but that still takes time, the effects of degradation of quality in software are immediate. Of course, it’s very difficult to dislodge a dominant software product from its place in the market, because people are willing to tolerate a lot of quality degradation if the cost of switching to something else is high, but there is an upper limit of what people are willing to take, while there is no limit to how bad their software can get if companies keep riding the LLM bandwagon.