More open engineering roles than last year. More available senior engineers than in recent memory. And neither side can find the other. A look at the 2026 job market data from TrueUp, Workforce.ai and Indeed, and why the bottleneck isn't supply or demand but verification.
Ask an engineering leader how hiring is going and you'll hear that good people are impossible to find.
Ask an experienced engineer how the job search is going and you'll hear that applications disappear into a void.
Both groups are describing the same market. Both are telling the truth. And according to The Pragmatic Engineer's 2026 job market deepdive, this standoff has now persisted for two straight years: hiring managers reporting it's hard to find engineers, job seekers reporting it's hard to get a response.
That's a strange thing for a market to do. Markets are supposed to clear.
The obvious answer would be that demand collapsed. The data says otherwise.
Per TrueUp, which tracks openings at top-paying tech companies, software engineering recruitment has climbed steadily since March 2023, with roughly 20% more roles posted than a year ago. Indeed/FRED data shows US and UK listings up over the past twelve months. Google has advertised around 62% more engineering roles than last year. Apple, IBM, and Amazon remain the largest sources of open positions.
Growth is even sharper outside Big Tech. Over two years, engineering headcount grew roughly 94% at Ramp, 84% at Wiz, 68% at Datadog, 55% at Rippling, 41% at Figma, and 37% at Netflix.
The other obvious answer would be that supply collapsed. That's not it either. Layoffs at Meta, Oracle, Atlassian, and Snap have put a large number of experienced engineers into the market this year: the same people hiring managers say they can't find.
So there are more open roles than last year, and more available senior engineers than in recent memory. And the two sides still can't locate each other.
When a market is liquid on both sides and still fails to clear, the problem isn't volume. It's verification.
Our read: this is the downstream cost of the collapsed hiring signal.
Last week this newsletter covered the data on that directly. Fabric, an AI interview platform, flagged 48% of engineering candidates for undisclosed AI assistance across roughly 19,000 interviews, and 61% of those flagged still passed. In a separate interviewing.io survey, 81% of interviewers said they suspect candidates of using AI, while only around 11% of companies use any detection at all.
Take those two facts together and you get a market where employers no longer trust what their own process tells them.
What follows is rational on both sides, and worth being fair about. If you can't trust your screening, you raise the bar, add rounds, and slow down. Every hire feels like a coin flip with six figures on it. If you're a strong engineer whose applications keep vanishing into that same widened funnel, you apply more broadly. Neither response is wrong. Both are individually sensible.
Collectively, they're a disaster. Employers screen harder, so throughput drops. Candidates apply wider, so volume rises. The ratio of noise to signal gets worse every quarter, and both sides work harder for a worse outcome.
That's not a talent shortage. It's a verification shortage. And it doesn't fix itself with a better economy, because the economy isn't what broke.
Two other patterns in the data make the mismatch worse.
The demand moved. AI engineering listings are up 50-100% at many larger tech companies while mobile and frontend demand has softened. And the entry-level door remains largely shut: the same reporting notes it's still harder for new grads and interns to get hired even as overall hiring rises, which squares with what this newsletter covered in May.
The volatility is also concentrated. Meta grew engineering headcount nearly 20% over two years, then cut 10%. Oracle announced up to 30,000 layoffs in March and dropped out of the top 20 by openings. When the most visible companies whipsaw, the mood follows them, even while less-visible companies are quietly growing 60-90%.
There's a timing wrinkle too. Workforce.ai's data suggests net hiring clusters between March and June, with little net growth in the second half of the year: companies set headcount budgets in January and largely spend them by mid-year. That's a pattern rather than a law, and it may not repeat. But if it holds, a req opened in late July is competing for attention against budgets that are mostly committed.
A large company can absorb a broken hiring process. It runs more rounds, eats the cost of mis-hires, and has a bench to cover the gap.
A ten-person team has none of that. Every hire is a meaningful percentage of engineering capacity. A mis-hire doesn't just underperform. It pulls your best engineers off their own work to review, rewrite, and hand-hold, which is the most expensive thing that can happen to a small team. And you're competing for exactly the specialists that well-funded companies are hiring fastest: infrastructure, security, observability.
So the small team faces the worst version of the problem. Highest stakes per hire, least margin for error, weakest position in the bidding, and the same unreliable signal as everyone else.
They stop treating every capacity gap as a headcount decision, and separate "we need this capability now" from "we need this person permanently."
They weight demonstrated production work over interview performance, because the interview signal degraded and the work signal didn't. Someone's track record on real systems under real conditions is harder to fake than a 40-minute session, and McKinsey's 2025 technology talent report found teams using structured, defense-based assessments reported 41% fewer early departures than those relying on live-coding rounds alone.
And they stop timing critical delivery to a hiring cycle they don't control.
The hardest part of hiring right now isn't finding people. It's verifying them.
That's the part we remove, and it's why this doesn't reset in January. The engineers we embed are senior people whose judgment has already been demonstrated on production systems: debugging without a trail, interrogating AI output, owning what they ship. The verification happened before they reach you, on real work rather than in an interview room.
So you're not running a broken screening process faster. You're skipping it. You see the work first and decide from evidence, which is what hiring was supposed to give you and currently doesn't.
The market has plenty of engineers and plenty of jobs. What it lost was the ability to tell which is which.
© 2026 Percime Technologies. All rights reserved.
