A software engineer who goes by “Matt” commutes four hours each way to Pawling, New York, using the time to code a video game entirely by hand. Not because he’s a hobbyist. Because he’s afraid of losing his skills. “I am actively trying to keep my axe sharp,” he told The Guardian. His job has shifted from writing code to reviewing what AI generates. His manager criticized him for not using AI enough. His future, he says, “feels dark.”
🔍 THE BOTTOM LINE
Google says 75% of its code is now written by AI. Over 600,000 US tech workers have lost their jobs since ChatGPT launched. Computer science enrollment is falling. The software engineering profession — one of the highest-paid, most stable careers of the past decade — is being restructured in real time, and the people inside it are terrified, angry, and scrambling to adapt.
A Profession Unraveling
The Guardian interviewed more than a dozen software engineers for a piece that reads less like a trend story and more like a dispatch from a profession in crisis. The details are specific and damning. Matt’s salary, typically over $200,000, used to feel like a sure thing. After a layoff last summer and a warning from his boss, he’s writing a game by hand on his commute to keep his skills from atrophying.
Another engineer, “Sam,” pivoted to software a decade ago after dropping out of a music degree and racking up $130,000 in student loans. Five years in, he’s gone from feeling professionally stable to wondering if he should open a food truck or get into forestry. “AI has taken over the creative, fun part of the job,” he said, “and reduced it down to the worst part: reviewing code I didn’t write.”
The numbers back up the anxiety. Nearly 50 million people worldwide worked as developers in 2025, according to SlashData. But since ChatGPT’s release in late 2022, more than 600,000 US tech workers have been laid off. The unemployment rate for computer science graduates rose to 7% in 2024, with underemployment above 19%. US tech job postings on Indeed dropped 36% from 2020 to 2025.
This isn’t a slow evolution. It’s a rapid restructuring that caught the workforce off guard. As we reported when 92,000 tech jobs were cut in Q1 2026, the pace is accelerating, not stabilizing.
The Adaptation Spectrum
Engineers are responding in three distinct ways. Some are doubling down on fundamentals — going back to basics, learning to evaluate AI-generated code for vulnerabilities, errors, and security issues that non-coders can’t spot. George Dover, laid off from Mailchimp in late 2024, spent his unemployment using AI to build websites and then rigorously auditing the output. After 400 applications, he landed a job oriented toward AI.
Others are organizing. Kaitlin Cort quit software engineering to found What We Will, a resource center for tech workers affected by AI disruption. She receives at least 10 new applications daily and says requests for unionization help have surged in recent months. The center has run campaigns targeting Amazon, Oracle, and Meta, helping workers with severance negotiation, benefits access, and organizing. “We just, as an industry, don’t have a guild,” Cort said. “We don’t have regulations or standards that are really shared.”
A third group is contemplating exit entirely. Sam’s food truck fantasy isn’t unique. Enrollment in computer and information science programs at four-year universities fell 8.1% in the 2025-2026 school year, with graduate enrollment dropping 14%. “Learn to code” — the mantra of the 2010s, championed by everyone from Barack Obama to Mark Zuckerberg — is starting to look like advice from a different century.
What the Experts Say
Bouke Klein Teeselink, assistant professor of economics at King’s College London, puts it bluntly: “The skill of writing code is over.” But he adds a crucial qualifier — the ability to evaluate AI-written code is becoming more important, not less. “AI is hugely augmenting what it means to be a software engineer.”
Ethan Mollick of Penn’s Wharton School frames the shift as a skills rotation rather than a replacement. “Now it’s not about who can write the most code. It’s about defining problems, designing systems and directing AI tools effectively.” The value has moved up the stack — from implementation to architecture and judgment.
Harvard’s David Malan offers a cost-based counterargument. OpenAI reportedly spent $8 billion on compute, and Anthropic is expected to have burned $3 billion last year. Those costs will eventually pass to customers. Rather than companies going fully AI, Malan expects “a healthier balance of software engineers being supported by AI.” This aligns with what we saw when Cloudflare cut 1,100 jobs and replaced them with AI agents — the economics of full replacement are still uncertain at scale.
The NZ Angle
New Zealand’s tech sector is smaller and more concentrated than the US market, but the dynamics are similar. Local firms have been adopting AI coding tools aggressively. The risk for Kiwi engineers is more acute in one sense: the NZ tech market has fewer employers to pivot to if you’re laid off, and the geography makes remote work for US companies the primary fallback. The unionization trend Cort describes is almost entirely absent in NZ tech, where collective action has historically been weak. Engineers here face the same disruption with fewer safety nets.
The Deeper Question
The Guardian piece surfaces a tension that the industry hasn’t resolved: if the value of coding skills is declining but the need to evaluate AI-written code is rising, who actually benefits? The engineers who can audit AI output are the same ones who spent years learning to write code — the very skill now being devalued. The pipeline that produces auditors (learning to code) is the same pipeline that’s shrinking (CS enrollment dropping). There’s a lag between the devaluation of writing code and the demand for evaluating code, and engineers are caught in that gap right now.
As we noted when Gallup found half of US workers fear AI job displacement, the fear is not irrational. It’s early. And for the engineers profiled in the Guardian piece, it’s already here.
❓ FAQ
Is software engineering dead? No, but it’s restructuring. The ability to write code is becoming less valuable; the ability to evaluate, architect, and direct AI tools is becoming more valuable. Engineers who adapt are still in demand. Those who don’t are vulnerable.
How many tech workers have been laid off since ChatGPT? Over 600,000 in the US alone, according to Layoff.fyi. That’s across all tech roles, not just software engineering, but engineering has been heavily affected.
What should current software engineers do? The experts in the Guardian piece converge on one answer: learn to evaluate AI-generated code rigorously. Security, vulnerabilities, performance, edge cases — these are the skills that non-coders can’t bring. The engineer’s value is shifting from production to judgment.
Is CS enrollment really dropping? Yes. The National Student Clearinghouse reports 8.1% enrollment decline in undergraduate CS programs and 14% in graduate programs for the 2025-2026 school year. After a decade of “learn to code” messaging, the next generation is voting with their feet.
What is What We Will? A resource center founded by former software engineer Kaitlin Cort in February 2026, helping tech workers navigate layoffs, negotiate severance, upskill, and organize unions. It has run campaigns with Amazon, Oracle, and Meta workers.
🔍 THE BOTTOM LINE
The Guardian’s reporting confirms what the layoff numbers have been signaling for months: the software engineering profession is going through a structural break, not a temporary downturn. The engineers who survive it will be the ones who can do what AI can’t — judge the quality, security, and correctness of generated code. The ones who can’t make that shift are looking at food trucks and forestry. The decade of “learn to code” is over. What replaces it hasn’t been named yet.
