METR study finds AI is slowing down software engineers by nearly 20%, not speeding them up
The issue started last week when METR, a small AI research nonprofit, published a surprising report saying software engineers took 19% longer to finish coding tasks with AI tools. The slowdown contradicts what developers expected — they had predicted AI would boost speed by 24%.
Developers even misjudged the outcomes afterward, still believing AI sped them up 20%, despite taking longer. 
“When developers are allowed to use AI tools, they take 19% longer to complete issues — a significant slowdown that goes against developer beliefs and expert forecasts,” METR said.
“This gap between perception and reality is striking: developers expected AI to speed them up by 24%, and even after experiencing the slowdown, they still believed AI had sped them up by 20%.”
The results expose limits in current AI coding tools during real-world use.
The slowdown hit amid heated AI startup deals. On Monday, AI coding company Windsurf was snapped up by Cognition after a failed deal with OpenAI. Google poached Windsurf’s CEO and scored a $2.4 billion license deal. Meanwhile, Cursor raised $900 million in May, valuing the company at $10 billion. The hype around “Vibe coding” — writing code entirely relying on AI — is booming.
AI engineer pay is skyrocketing. Meta is offering multimillion-dollar packages for AI researchers. LinkedIn says “AI engineer” is the fastest growing title among recent grads, outpacing traditional software jobs.
Still, overall software developer job openings hit a five-year low, fueling worries AI is replacing coders. Microsoft cut 2,000 jobs recently, mostly software engineers, despite CEO Satya Nadella stating up to 30% of their code is now AI-generated.
MIT researchers dropped a new paper outlining why AI still struggles with complex and large-scale coding. Senior author Armando Solar-Lezama said:
“Everyone is talking about how we don’t need programmers anymore, and there’s all this automation now available,”
“On the one hand, the field has made tremendous progress. We have tools that are way more powerful than any we’ve seen before. But there’s also a long way to go toward really getting the full promise of automation that we would expect.”
Experts say the coder job crunch may have more to do with economic pressures than AI itself. Heather Doshay from SignalFire put it bluntly:
“Teams are getting smaller. Not necessarily because of AI, but because of market demands and operating expenses. What’s happening is companies are asking, ‘How can we stay lean and hire fewer people while still extending our runway financially?’”
Anxiety is high among coders. Layoff tracking sites show tech job cuts rising for three straight quarters, though not near 2023 peaks. On Blind, devs debate if AI is actually killing jobs or if the trend is a cover for business cuts.
One software engineer at a tax firm described the tightening market:
“The expectations for an engineer are way up. We’re now only seeing the top talent get hired. It’s intimidating.”
The AI code race is speeding up but for now, it’s slowing down developers.
