AI Layoffs Are Real — But So Is the Hype: The Automation Reality Check Beginners Need
The wave of AI layoffs in 2026 has already hit hard — 115,430 people lost their jobs across 152 tech companies in just the first five months. That number nearly matches the 124,636 layoffs from all of 2025 combined. The common thread? Companies are blaming AI. But here’s the thing nobody’s telling you: the research behind these decisions is a lot messier than the headlines suggest. If you’re worried about AI replacing your job — or if you’re excited about automating your way to a higher income — you need to understand what’s actually happening.
TL;DR — The Honest Summary
- Over 115,000 tech workers were laid off in the first 5 months of 2026, with most companies citing AI as the reason
- ClickUp cut 22% of its staff and replaced them with roughly 3,000 AI agents, promising “million-dollar salary bands” for remaining employees
- A respected research lab (METR) found that AI actually slowed down experienced developers by 20% in controlled tests
- AI-generated code has 1.7x more bugs than human-written code, and companies spend 44% of their AI tokens just fixing problems the AI created
- Cognition raised $1 billion for its AI coding agent Devin, but its own CEO says it should “augment, not replace” humans
AI Layoffs in 2026: What the Numbers Actually Say
115,000+ People Laid Off in 5 Months — and AI Gets the Blame
The layoff wave of 2026 is unlike anything we’ve seen in recent tech history. According to Layoffs.fyi, 152 companies have already cut more than 115,000 jobs this year. By comparison, the entire year of 2025 saw 275 companies lay off 124,636 people. In other words, we’re on track to double last year’s total in half the time.
However, the story behind these numbers is more complicated than it appears. Gartner found that roughly 80% of companies using autonomous technology have cut jobs. Yet the same research suggests those workforce reductions aren’t necessarily translating into meaningful financial returns. Meanwhile, UC Berkeley published a meta-analysis in October 2025 concluding there is “no robust relationship between AI adoption and aggregate productivity gain.” Similarly, NBER research from March 2026 found what it calls a “productivity paradox” — perceived gains consistently outpace measured gains.
What the Research Says About AI Productivity
MIT researchers evaluated thousands of worker tasks and concluded that AI agents still can’t reliably do human-quality work across most domains. They predict models might reach 80-95% success at “minimally sufficient quality” by 2029, with several more years beyond that before they could consistently outperform humans.
Harvard Business Review added another layer in May 2026: when everyone uses AI to produce more output, the bottleneck shifts to the managers who have to review and approve all of it. More output doesn’t automatically mean more value.
The ClickUp Story — 22% Laid Off for AI Agents
What Really Happened at ClickUp
In late May 2026, ClickUp CEO Zeb Evans announced that the company was laying off 22% of its workforce. But he framed it differently than a typical cost-cutting move. According to Evans, this was a “radical embrace of AI,” not a layoff in the traditional sense.
ClickUp had deployed roughly 3,000 internal AI agents to handle tasks previously done by human employees. Evans described his vision as building a “100x organization” — where the remaining staff would direct these agents and review their output rather than doing the work themselves. Furthermore, he promised “million-dollar salary bands” for people who create outsized impact using AI.
On X, Evans wrote: “The people that automate their jobs with AI will always have a job.” The message was clear: adapt or leave.
Aaron Levie’s Warning: “AI Psychosis” in the C-Suite
Aaron Levie, the CEO of Box and a well-known AI proponent, coined a term that cuts right to the heart of the problem: “AI psychosis.” Levie argues that tech CEOs are uniquely vulnerable to overestimating AI’s capabilities because they’re too far removed from the actual day-to-day work. They see polished demos and prototype a contract, then leap to believing agents can do the entire job.
As Levie put it on X: “CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI.”
For beginners trying to understand the AI layoff wave, this distinction matters. The people making the hiring and firing decisions aren’t always the ones who understand what the AI can and can’t do. You can read our full breakdown of What Is AI Psychosis? (And Why Your Boss Might Have It) to go deeper on this topic.
Is AI Really Replacing Jobs? The Research Says Slow Down
METR Study: Developers Refuse to Work Without AI — But It May Not Help
METR, a respected AI safety research lab, ran a study in 2025 that produced a shocking result: AI tools actually caused a 20% slowdown in task completion among experienced open-source developers. When METR tried to repeat the experiment with updated tools later that year, developers refused to participate. They didn’t want to work without AI — even when offered $50 per hour on tasks of their own choosing.
Furthermore, 30 to 50% of developers admitted to selectively avoiding tasks they thought AI handled well, because they didn’t want to risk being assigned to the “no AI” group. As one developer put it: “My head’s going to explode if I try to do too much the old fashioned way because it’s like trying to get across the city walking when all of a sudden I was more used to taking an Uber.”
METR’s follow-up survey of 349 technical workers in May 2026 did show strong self-reported gains — a median 3x speed improvement and 1.4-2x value uplift. But the researchers included a crucial caveat: their earlier controlled study found that people overestimated AI’s time savings by an average of 40 percentage points. Perception and reality are not the same thing.
AI Code Has 1.7x More Problems (CodeRabbit)
The quality question gets worse. CodeRabbit, a code review tool, analyzed open-source pull requests and found that AI-generated code has 1.7x more problems than human-written code. Meanwhile, Entelligence AI’s CEO reported that companies spend 44% of their AI tokens just fixing bugs that the AI itself created.
A large-scale study from Singapore Management University backed this up with hard data. Researchers analyzed 302,600 AI-authored commits across 6,299 GitHub repositories and found 484,366 distinct issues introduced by AI. Moreover, 22.7% of those AI-introduced issues still existed in the latest version of those repositories — meaning they became permanent technical debt.
The Tokenmaxxing Problem — More Code, More Debt
Here’s where the story gets particularly relevant for anyone using AI tools at work. “Tokenmaxxing” is the practice of measuring developer productivity by how many AI tokens they consume. It sounds reasonable in theory. In practice, it’s a disaster.
Amazon shut down its internal AI leaderboard called Kirorank after employees gamed the system by running up token usage without producing meaningful work. Uber, meanwhile, blew through its entire 2026 AI budget in just four months. According to Uber’s COO Andrew Macdonald, that spending “hadn’t led to a measurable increase in projects or productivity.”
Several analytics companies have documented the pattern. Jellyfish found that engineers with the largest token budgets produced twice the code at ten times the cost. GitClear reported 9.4x higher code churn among AI users. Faros AI found an 861% increase in code churn under high AI adoption. We covered the token billing side of this in detail in our post about GitHub Copilot Token Billing.
The Billion-Dollar Irony: Cognition Raises $1B, but Says Don’t Replace Humans
Scott Wu: Devin Is Between Junior and Mid-Level
On May 27, 2026, Cognition — the company behind the Devin AI coding agent — raised $1 billion at a $26 billion valuation. It’s the kind of number that makes headlines and fuels the “AI is replacing everyone” narrative. But listen to what Cognition’s own CEO actually says.
Scott Wu, a former child prodigy who started winning math competitions as a second grader, describes Devin as sitting “somewhere between a junior and a mid-level engineer.” He’s explicit about the limits: Devin should augment human programmers, not replace them. As Wu told TechCrunch: “We’ve never thought about it as replacing humans. It has never been our view.”
For a deeper look at Cognition’s approach, check out our article Cognition’s Devin Just Raised $1B — Here’s Why AI Coding Agents Won’t Replace You.
89% of Cognition’s Code Is Devin-Committed
Here’s the number that sounds scary at first: 89% of code committed by Cognition’s engineers was actually committed by Devin. The remaining 11% came from local agents in Windsurf, an AI coding IDE that Cognition acquired in July 2025.
But context matters. Devin primarily handles long-tail maintenance work — updating old code, migrating platforms, and tackling tasks that many developers find tedious. Cognition isn’t using Devin to design systems or make architectural decisions. Instead, it’s using AI for the kind of work that frees up human engineers to focus on what they’re actually good at. That’s augmentation, not replacement.
The irony is impossible to miss. The company that just raised a billion dollars to build AI agents is telling the world, in no uncertain terms, that humans still matter.
What This Means for Beginners — Practical Takeaways
So, what should you actually take away from all of this? Here are the key lessons:
First, the layoffs are real, but the reasoning behind them is flawed. Companies are cutting staff based on AI hype, not proven productivity gains. The research simply doesn’t support the narrative that AI makes workers dramatically more productive — at least not yet.
Second, AI is a powerful tool, not a replacement for human judgment. The data consistently shows that AI works best when humans are in the loop — reviewing, correcting, and directing. The companies getting the best results use AI to handle repetitive tasks while keeping humans focused on decisions and creative work.
Third, quality matters more than speed. Programmer James Shore published a viral post in May 2026 that cuts to the core issue. If AI doubles your output but doesn’t cut your maintenance costs, you’re “trading a temporary speed boost for permanent indenture.” The code you generate today becomes your problem tomorrow.
The Smart Way to Use AI Without Betting Your Job on It
If you’re just getting started with AI tools, here’s how to use them wisely:
- Use AI for tasks it’s genuinely good at — drafting emails, summarizing documents, brainstorming ideas, and automating repetitive workflows. Don’t use it as a crutch for decisions that need human judgment.
- Always review AI output before using it. The data is clear: AI introduces more bugs, more errors, and more long-term debt. Treat AI output as a first draft, not a finished product.
- Measure actual results, not just activity. The tokenmaxxing lesson applies to everyone, not just big tech companies. More AI output doesn’t equal more value. Track whether the AI is actually saving you time or just creating more work to clean up.
- Build skills AI can’t easily replicate. Critical thinking, domain expertise, creative problem-solving, and the ability to communicate clearly — these are the skills that get more valuable as AI becomes more common, not less.
- Stay informed but don’t panic. AI layoffs in 2026 are real, but they’re driven more by hype than by proven economics. MIT predicts AI won’t reliably match human-quality work until around 2029 at the earliest. That gives you time to adapt thoughtfully.
If you want to start building practical AI skills the right way, our Vibe Coding Career Guide: How to Start Coding with AI in 2026 (Without Becoming Dependent) walks you through the basics.
AI is changing the workplace — that much is true. But the smartest move right now isn’t to panic about AI layoffs 2026 or to go all-in on automation without understanding the trade-offs. Instead, the winners will be the people who learn to use AI as a tool while building the human skills that AI can’t touch.
—
