I’ve been through one or two dry spells. The early ’90s were a pretty difficult time for someone new to the industry. The dotcom crisis also caused the industry to tighten its belt. There has always been an ebb and flow in demand, but those who eat and breathe software sailed those seas.
This time it is fundamentally different. Life has been pretty good for software developers over the past 40 years, but life owes software developers nothing, and maybe we’re nearing the end of the gravy train.
Until 1900, the horse was the main means of transportation. In about ten years, the automobile replaced horses in volume. Horses were no longer needed. Sure, you can still find them today, perhaps at horse races or at tourist companies. But there is no law of nature that says “there will always be new jobs for horses.”
And I fear the same with programmers. Artificial Intelligence has arrived, it works at a very low cost compared to humans and can already perform much of the intellectual work of software developers. This is not to say that it can replace developers, at least not yet, but the trend is unmistakable.
The answer they tell us is to embrace AI, become data engineers, data scientists, and embrace all the shiny new AI technologies. In response to this call, there have been broadly two camps in software development. The first camp prides itself on its intellect and condemns the use of LLM. For these developers, using AI is admitting their incompetence. Or maybe they believe a machine will never be able to replace them.
The second field is full of AI adoption. In a sense, I am in that field, as I have always been fascinated by AI and it has been a dream to work in that field. That said, there is a more pragmatic reason; that cutting-edge AI is truly a force multiplier. It allows me to achieve much more than without it. AI fans may hear exhortations to learn about AI to stay relevant in the next revolution.
However, the reality is that the pie itself will shrink. We have already seen companies lay off software developers, more in anticipation of the success of AI to replace them than for any compelling reason. At least until now.
Today’s LLMs are good. Five years ago, the LLMs we see today would be considered science fiction. But despite the successes of LLMs, they are still not capable of completely replacing humans. You still need humans in the loop. They can be a force multiplier for a software developer, allowing them to be much more productive than they would otherwise be. But now there are also ‘vibecoding’ platforms where people without development skills can develop software.
The AI revolution is not part of a cycle. There is a fundamental change taking place regarding artificial intelligence. In my naivety at first I believed that technologists like me would be the last to feel the pain of being replaced, but in reality it seems that we are one of the first along with artists, writers and musicians.
“Learn to code,” they said.
Although AI will not directly replace developers initially, it will reduce the number needed. An increasingly smaller number of developers will be needed. This will affect salaries as the supply of developers will exceed demand.
None of this requires anything as grand as magical AGI or even super intelligence. All that is needed is gradual improvement. They will be like musical chairs for software developers, with fewer and fewer chairs. To obtain a professorship, the demands in terms of skills and experience will only increase. This in itself will cut off the supply of younger, less experienced developers.
What should a software developer do? Should we do whatever it takes to find a chair? Or are we horses in the age of cars?
My advice is this: learn to be a hairdresser. Even with