It's not AI. It's the economy.

Four hundred applications. Maybe nine replies. Two of them were rejections and the rest were silence.

That's the experience a lot of very competent developers are having right now, and there's a ready-made explanation waiting for them: AI took the job. It's clean, it's dramatic, it explains everything, and it happens to be the story most convenient to the people doing the cutting.

I think it's mostly wrong. Not because AI does nothing — it does plenty — but because the thing actually crushing the job market is more boring, better documented, and much less flattering to management: the Western economy is in bad shape, and hiring is the first thing that freezes.

The timeline doesn't work#

Start with the sequence, because causes are supposed to precede effects.

The mass tech layoffs began in 2022. By the trackers' count, roughly 165,000 that year, then a peak of about 263,000 in 2023 — the worst year on record. (layoffs.fyi is the usual reference; the various trackers disagree by large margins depending on what they count, so treat all these numbers as directional.)

Now recall the state of the art in 2022. The models of that moment could produce a plausible function and hallucinate an API that didn't exist. Nobody — including the people running those companies — believed a model was replacing an engineering team. The layoffs were happening anyway, and the reasons given at the time were the honest ones: rate shock, then over-hiring correction.

What actually changed in 2022 was the price of money. A decade of near-zero rates ended, and software is the most rate-sensitive asset there is — it's a bet on cash flows years out, financed today. When money got expensive, the bet got expensive, and headcount is the largest line item in that bet.

There was a second, quieter hit that almost nobody outside finance talks about. From the 2022 tax year, US companies could no longer immediately deduct software development salaries — Section 174 forced them to amortize those costs over five years. The after-tax cost of employing a developer went up sharply, for accounting reasons, at the exact moment rates spiked. Congress finally restored immediate expensing for domestic R&D in the 2025 tax bill, effective for tax years starting after 2024, which tells you how seriously the damage was eventually taken.

None of that is artificial intelligence. All of it lands on the same budget line.

The data says "frozen," not "replaced"#

Here's the detail that changed how I read this whole situation.

If AI were substituting for developers, you'd expect firings: companies dropping the humans they no longer need. That's not what the labor data shows. The US labor market in 2026 is a low-hire, low-fire freeze. The June JOLTS numbers had the hires rate around 3.4% and layoffs near 1.1% — and by July the hires rate slipped to 3.2%, the weakest hiring since 2010, when unemployment was near 10%.

Sit with that. Hiring at crisis-era lows, while actual layoffs sit at historically low levels.

That is not the signature of an industry replacing workers with machines. It's the signature of an industry that has stopped opening positions and is waiting to see what happens. Information-sector openings are down roughly a third year over year, and the quits rate in that sector has fallen by nearly half over four years — nobody is leaving voluntarily, so nothing backfills, so the ladder stops moving. The market isn't dead. It's jammed.

The people who study this for a living are skeptical#

I'm not the one making this argument. It's the mainstream position among the economists watching the numbers.

Ben May, who runs global macro research at Oxford Economics, put it about as bluntly as an economist will: firms are trying to "dress up layoffs as a good news story rather than a bad one — for example, by pointing to technological change instead of past overhiring." Their broader finding is that evidence of an AI-driven shakeout is patchy, and that most employers don't appear to be replacing meaningful numbers of workers with AI. Other analysts have reached the same conclusion: the cuts track weak demand and excess hiring far better than they track any deployed capability. The practice even has a name now — AI washing — and Sam Altman is among the people using it.

And then there's the arithmetic that gives it away. The largest US tech companies are on track to spend something in the neighborhood of $700 billion on AI infrastructure this year, while cutting tens of thousands of staff and citing AI as the reason.

Read that again. The AI didn't eliminate the payroll. The payroll is paying for the AI. Datacenters are expensive, the capex was promised to shareholders, and it has to come out of something. "We're becoming more efficient with AI" is a much better line on an earnings call than "we committed to a spend we have to fund by cutting your team."

Europe deserves its own paragraph#

If you're job hunting in Europe and it feels worse than the noise online suggests, you're not imagining it.

The eurozone is looking at growth under 1% this year. Germany — the industrial engine of the continent — went through two years of stagnation and came out near zero. EU private-sector investment in innovation runs at roughly half the US level as a share of GDP, and the Draghi report put the annual investment gap at around €800 billion and warned of a "slow agony" of decline without serious reform. Two years later, a small fraction of its recommendations have been implemented.

European developers did not get worse at their jobs in 2023. The capital that used to hire them stopped showing up, and the structural reasons for that were diagnosed in detail, in public, by a former head of the ECB, and mostly ignored.

What I'm not saying#

I'm not saying AI is irrelevant to any of this. Three things are true at once, and pretending otherwise is how you end up with a comforting story that's just as useless as the scary one:

AI genuinely compresses junior-shaped work. The starter tasks that used to justify a first hire are the ones models handle best. That's real, I've written about it before, and it's why entry is harder even in a healthy market.

The stated reason still has real consequences. When executives believe — or claim to believe — that AI absorbs the work, the position doesn't get opened. The belief moves the budget whether or not the capability is there. A job you didn't get for a fake reason is still a job you didn't get.

And 2026 really is running hotter on cuts, with AI cited more and more often. Challenger's tracking has AI as one of the top stated causes in recent months. Stated. That word is doing a lot of work, and the analysts above are exactly why.

Why the diagnosis matters#

Because the two stories imply opposite behavior, and one of them is wrong.

If the profession is being deleted, then studying is sunk cost, your experience is a depreciating asset, and the rational move is the exit — retrain into something else immediately and stop wasting years.

If it's a capital cycle, then this is a bad stretch inside a career that lasts decades, and the rational move is to stay positioned. Cycles turn. Rates come down. Section 174 already got fixed. The frozen market thaws the moment quits pick up, because one person moving opens a chain of backfills behind them. And the enormous infrastructure buildout being used to justify today's cuts still has to be staffed by someone — that's the argument I made in the last post on this, and the capex numbers above are the strongest evidence for it.

The diagnosis also changes where you look. If the problem is Western capital sitting on its hands, then go where the capital is actually moving: infrastructure, energy, defense, healthcare, industrial software, the Gulf, Asia, LatAm — and away from the fifteenth consumer SaaS in a saturated market, which is exactly the category that only existed because money was free.

And it changes how you read the silence. Four hundred applications with nine replies is a fact about market structure — about openings that were never really open, pipelines with no backfill, and a hiring manager whose req got frozen in Q2. It is not a verdict on your worth as an engineer. Those feel identical from the inside, and they are not the same thing.

The practical move#

Stop reading the layoff-a-day content. It's optimized to make you feel exactly this and it will not tell you anything actionable.

Read the macro instead — rates, capex, where investment is actually flowing. It's duller and it's the thing that's genuinely determining whether your application gets read.

Then act like someone in a frozen market rather than a dying one. Frozen markets reward the people who are still there, still sharp, and impossible to overlook when things unstick: keep shipping, keep your network warm, and get good at telling the story of your work, because when there are five openings instead of fifty, being the obvious choice is the entire game.

Software isn't dead. Money got expensive, Europe stalled, and a lot of executives found a more flattering thing to blame it on.

That's a much better problem to have — because that one ends.