22-06-2025
Will there be more software developers working next year?
For the first time, the number of software jobs in the US has stagnated. This coincides with the broader trend of lower labor force participation across sectors.
Previously, recent college grads always had a lower unemployment rate than the general workforce. That has flipped — now recent grads have a notably higher unemployment rate.
Societal shifts because of new technology tend to be more noticeable over a decade than over a single year. Is AI different?
Brian Brackeen says there will be fewer software developers working in the United States a year from now. I call bullshit.
We're making a bet of it: Loser has to wait in line to buy the other a cheesesteak from South Philadelphia's beloved Angelo's — since we'll both be in that city for a tech-inspired celebration of the 250th anniversary of the Declaration of Independence around the next Builders Conference (Here's coverage of our most recent one).
Contrary to what I say to my friend's face, Brackeen is no fool. The proven entrepreneur–turned–provocateur venture capitalist has a point.
Two big changes are hitting software development at the same time, making it difficult to distinguish between them: Higher interest rates have chilled speculative tech building, and new artificial intelligence tools are creating new efficiencies. Further complicating the trend, pandemic habits boosted international tech hiring, and the decades-old digital transformation appears to be aging. Expensive software developers seem an ideal role for executives to replace.
In 2024, the American economy added new software developer jobs at the slowest year-over-year rate on record.
All this comes in a strange macroeconomic moment.
Take the first quarter of 2025: Tech hiring didn't just stall — it retreated. Software developer job postings were still falling even long after the pandemic-fueled bonanza had faded, a Wall Street Journal analysis found. Tech unemployment climbed above the national average, peaking at 5.7% in February.
Other than robotics, most current AI excitement is concentrated in knowledge work, something college graduates specialize in. (The plumber I once worked for used to say that getting a college degree was a way to buy a chance to work in air-conditioning). So as economists research whether AI is affecting job prospects, they're focusing on degree holders.
A real change is underway. For decades, the unemployment rate for recent college graduates was almost always lower than the overall unemployment rate. Following the Great Recession, unemployment peaked in June 2010 at 9.5%; It was bad for recent college graduates too, but that rate never got above 7.8%, according to the Federal Reserve Bank. The same thing happened in the recessions of 1990 and 2001.
But those figures have flipped in recent years — and the gap is widening. The first quarter of 2025 ended with an unemployment rate for recent college graduates that was almost one-and-a-half times higher (5.8% versus 4% for all workers). Amazon CEO Andy Jassy told employees this week that over the next few years, the online commerce giant 'will reduce [its] total corporate workforce as we get efficiency gains from using AI extensively across the company.'
Any move toward the end of knowledge work bodes poorly for software development, an especially well-paid (and occasionally rote) trade.
Fewer people are working — but not (yet) because of AI
This coincides with a broader trend: Labor force participation remains down from pre-pandemic levels. A recent explainer from the Federal Reserve underlined the stakes: Either we grow our economy by improving productivity or by adding more workers. With aging demographics and early retirements, we've been trending in the opposite direction.
That drop isn't theoretical. There are 1.7 million fewer Americans in the workforce now than in February 2020, per the US Chamber of Commerce. That means many jobs, including tech roles, are going unfilled.
So software developers are harder to come by. But I'm still betting we'll have more of them by next year. Why?
For one, the trends don't line up with AI breakthroughs. The unemployment rate for recent college grads crested above the overall rate way back in 2018 – though the gulf has gotten larger. Something else is happening.
Next, pricing pressure can create surprising outcomes. Often called the Jevons Paradox, falling costs in something (like the price of building software) can boost demand for it (resulting in more need).
The time required for any given software development can shrink, even as the need for people to do adjacent work can grow. Job titles can change — witness the decline of ' computer programmer ' — and job descriptions too. But software skills will remain in-demand for the foreseeable future.
The third reason I took the bet is the maxim credited to Bill Gates: We tend to overestimate what we can achieve in a year and underestimate what we can in a decade. (Note: I was less sure about taking Brian up on a bet over 5 years, but jumped at the bet for a year out — check the video here.)
Early this year, University of Oxford researchers published a paper documenting the impact automation has had on language translators. It's changing the job and shifting the skills needed, but certain tasks remain stubbornly human.
Will AI be different from other tech advancements?
No question getting a job, especially a first job, in tech is harder than it has been.
The leader of one coding bootcamp told me there's been 'a collapse,' and they're feverishly adapting their model. At present, that is more to do with higher-interest rates and general macroeconomic trends, with AI as window dressing. Focused on the medium term, boosters remain bullish that the growth in tech apprenticeship will continue.
Brackeen's side of the bet is that AI is different from past technologies, and that the speed will sneak up on us. AI legend Geoffrey Hinton recently argued the same. My side of the bet is that even if this is right, it will happen slower than we think, and that titles like software developer will adapt, and therefore grow, for at least a while longer.
We'll settle it over cheesesteaks. See you in Philadelphia next May.
