Why This Can't Just Be Blamed on the Economic Cycle

Several macro shifts unrelated to AI really were happening over this same period. To confirm this collapse is mainly AI's doing, each of these suspects needs to be ruled out one by one.

Why the Aggregate Data Looks Fine

Zoom out to the whole economy, or even just "computer occupations," and you'll see a strange picture: every headline number is up.

From May 2024 to May 2025, total US employment grew 0.8%, while computer and mathematical occupations grew 1.3% — faster than the overall economy. By the Bureau of Labor Statistics' count, employed software developers grew from 1.53 million in May 2022 to 1.69 million in May 2025 — up 10% across the entire AI era. Rigorous studies out of the US, Denmark, and even Anthropic itself have found no relationship between AI exposure and overall employment; the Danish study, using government payroll ledgers, could rule out any effect larger than roughly 1%.

How can both things be true at once? Weight each age cohort by its share of the workforce, and the answer appears.

Same data, weighted by each age cohort's workforce share: total developer employment is up 4.4% since October 2022, while the 22-25 cohort is down 19%. Source: Stanford Digital Economy Lab / ACS PUMS weights

Which Specific Coding Work Is AI Actually Eating

Zoom in further to see which specific job titles are shrinking, and the picture gets even clearer. Same Bureau of Labor Statistics data, May 2024 to May 2025.

US employment change by occupation, May 2024 to May 2025. Contracting occupations on the left, growing ones on the right. Source: BLS Occupational Employment and Wage Statistics

The "computer programmer" occupation — which the Bureau of Labor Statistics defines as "someone who writes code to someone else's specifications" — fell 16% in a single year. The BLS's original forecast for this occupation was a decline of just 6% per decade. Web development, the author's own field, fell 11%; QA testing fell 6.5%. Meanwhile, data scientists grew 12%, systems analysts grew 4.4%, and the broader "software developer" category grew 2%.

The New Developers Really Did Show Up — Nobody Just Calls Themselves a Programmer

Back in 2025, the author wrote that AI is a new layer of abstraction, and like every abstraction layer before it, it would spawn far more developers building far more software. He also argued that these new people should be called "software developers" too — inventing a separate name would only manufacture an artificial barrier. He now thinks he was half right: the wave of new developers really did arrive. They just don't use that title.

This software boom is real, and measurable. In its last Octoverse year, GitHub added 36 million new accounts — the fastest growth rate in the platform's history, roughly one new developer every second — while also adding 121 million new repositories, the single biggest year of repo creation in the platform's history, so much that its infrastructure was creaking at the seams. Of these newcomers, 80% used Copilot within their first week. The largest developer influx in history is AI-native, and it happened at precisely the same moment paid entry-level jobs were collapsing.

The author's favorite piece of evidence is the App Store, because publishing an iOS app has real cost and a real barrier to entry: a $99 developer fee, a review process, and a finished product that actually works. It measures software that's genuinely shipped, not tutorial exercises.

New iOS App Store software submissions per year. Peaked in 2016, then declined for eight straight years; reversed in 2025 with 24% growth. Source: Appfigures

New App Store submissions peaked in 2016 and then declined for eight straight years. In 2025, they grew 24% — the first real growth since the peak — and by Q1 2026, iOS submissions were up another 80% year-over-year. The surge was big enough to stretch Apple's review times from two days to several weeks. The category mix shifted too, toward productivity, tools, and lifestyle apps — exactly what you'd expect from "people picking up coding for the first time to solve their own problem," not studios chasing game revenue.

Who are these people? According to Vercel, 63% of vibe-coding users are non-developers. Lovable says 60% of its users are "non-developers," and those users create over 100,000 new projects every day. Replit claims 50 million people have used its platform. These are marketers, founders, teachers, analysts, product managers — they're writing software, which in the author's view makes them developers. They just don't identify that way, and more importantly, it's not their job title — and job titles are exactly what labor statistics count.

The Broken Mentorship Chain

So, grading the author's 2025 prediction: right about the developers, wrong about the job title. That sounds like a happy ending — until you ask what comes next.

The traditional entry point into professional software engineering worked like this: you get hired to write mediocre code, a senior engineer reviews your code, you absorb judgment through repeated rounds of correction, and ten years later, you're the senior doing the reviewing. That chain is now broken. AI now writes the mediocre code, so nobody hires juniors, so nobody is in line to become the senior engineer doing the reviewing.

Meanwhile, millions of new builders are shipping software, and no one is reviewing any of it. A Veracode study found that 45% of AI-generated code fails basic OWASP security tests (the OWASP baseline is an industry-standard checklist for catching the most common, most exploitable software vulnerabilities). A separate audit of vibe-coded apps found that 10% had serious row-level security flaws actively exposing user data (a row-level security flaw means database permissions are misconfigured, so users who should only see their own data can read someone else's). Apple, meanwhile, is drowning in submissions it can't review fast enough. Software is being built, but the judgment layer hasn't kept pace with it — and the mechanism that used to build that judgment, the master-apprentice relationship inside employment, has collapsed.

Two Possible Futures: IBM Is Doubling Down, Salesforce Is Going to Zero

Even on this scorched earth of junior-developer jobs, a few hopeful shoots are appearing. Facing the same broken chain, two major companies have taken opposite paths.

These are the two candidate futures. Which one wins determines whether this industry still has senior engineers by 2036.

Has the Turning Point Already Begun

A market that refuses to hire juniors has one logical endpoint: the pain becomes real enough that it self-corrects. And that, maybe, just maybe, is already starting to happen.

Indeed's job-posting data actually bottomed out in May 2025, then rose for 13 straight months, up 10% year-over-year.

US software developer job postings on Indeed, February 2020 = 100. Peaked in early 2022, bottomed at 62 in May 2025, then rose for 13 straight months back to 72. Source: Indeed Hiring Lab

The author says that if Stanford's next update shows the 22-25 employment curve also turning positive, it may mean the market has found a new equilibrium. He suggests watching for more big employers launching programs like IBM's — and if none show up, someone will have to create them, or this whole software-creation boom will eventually tip into a bust. His summary: what we're watching is programming stop being a job title and become a skill, the same way "typist" stopped being a job and became something everyone is just assumed to know how to do. For almost everyone, that transition is going fine — except for the people who were just about to step onto the old ladder, right as we set it on fire.

AI now writes the mediocre code, so nobody hires juniors, so nobody is in line to become the senior engineer doing the reviewing.Laurie Voss · Seldo.com

Peel back the layers — by age, then by specific job title — and the picture becomes clear: what AI is eating isn't "programming." It's one very specific kind of work.

Here's the twist: the author's 2025 prediction that "AI will spawn a wave of new programmers" actually came true — GitHub added 36 million new accounts in a year (one every second), and App Store new-software submissions rebounded 24% after eight straight years of decline. It's just that these newcomers are product managers, teachers, marketers — they use AI to write software but don't call themselves programmers, and statistically get folded into their existing job titles, invisible in the official data. What collapsed is the job title "programmer"; the act of writing software itself is exploding. But that surfaces a bigger problem —

In the old model, one person climbs up rung by rung: write code → get reviewed by a senior → eventually become senior. Now AI has directly erased the bottom rung — XiaoHu can't even get an entry-level job.

Facing the broken chain, two companies gave opposite answers: IBM tripled its entry-level hiring, redesigning the role around "client-facing work + writing requirements" instead of pure typing; Salesforce hired zero engineers last fiscal year. Whichever path wins determines whether this industry still has senior engineers by 2036.

The US Bureau of Labor Statistics originally forecast the "computer programmer" occupation would shrink just 6% per decade. Instead, one year into the AI era, it dropped 16% — a decline originally projected to take 27 years happened in one.