In 2026, you can’t pry AI coding tools out of developers’ grip, researchers have found.
But while AI is undoubtedly helping coders produce code faster, it may not produce better code, warn other researchers. And that could cause them problems down the road.
Specifically, in February 2026, the AI METR research lab was respected posted a startling revelation: Most developers will no longer function, even on a limited number of tasks, without AI.
METR hoped to provide an update to some groundbreaking research published a few months earlier in 2025 on AI coding productivity. In it, the researchers measured how long it took open source developers to do tasks by hand compared to artificial intelligence.
While the programmers in this study reported that AI made them more productive, they were shocked to learn that it actually slowed them down. Sure, it generated code faster, but then they spent extra time finding and fixing bugs, directing the AI, and waiting for tasks to complete.
When METR set out to repeat the experiment to measure advances in AI and coding capability, it failed.
The developers were reluctant to participate “because they don’t want to work without artificial intelligence” even just for the study, the researchers confessed.
Instead, METR published research in May that allowed technical employees to self-report the productivity gains of artificial intelligence. Not surprisingly, they realized that AI made them twice as valuable to their organizations.
But recent headlines about the wild costs of so-called tokenmaxxing, combined with a minor recent survey, make those self-perceptions questionable.
Tokenmaxxing, or using the number of tokens a person uses as a proxy for productivity with AI, has been the trend of 2026 so far. And it may already be over.
Amazon shut down its internal token leaderboard called Kirorank after employees were gaming it by overusing AI agents and driving up costs. the Financial Times reported this week. Workers have proven that the use of artificial intelligence does not automatically translate into increased productivity.
Uber blew past its 2026 AI budget in the first four months of the year, The Information was mentioned. COO Andrew Macdonald recently said on a podcast that such spending had not led to measurable growth in projects or productivity.
AI-generated code also doesn’t necessarily reduce ongoing code maintenance needs, and may even increase it, developer and author James Shore has elegantly argued. a blog post that went viral on Hacker News.
“Write code twice as fast now? I better hope you’ve cut your maintenance costs in half,” he wrote. “Otherwise, you’re screwed. You’re trading a temporary speed boost for a permanent guarantee.”
There are other indications that AI may increase code maintenance problems.
A viral tweets by Aiswarya Sankar, founder and CEO of reliability engineering startup Entelligence AI, proclaims that companies spend 44% of their tokens on bug fixes created by their AI. Meanwhile, code review tools company CodeRabbit says it analyzed open source pull requests and found that AI generated 1.7 times more issues than human code.
These are, admittedly, self-serving statistics from those trying to sell AI code review tools.
However, independent researchers have also found such issues. Researchers from the esteemed Singapore Management University published a report in April warning that “AI-generated code can introduce long-term maintenance costs into real software projects.”
Since developers love their AI assistants, what’s the solution?
Well, those who want to sell you AI coding agents say that developers can simply use AI coding agents to do the tedious work of fixing code as fast as the AI spits it out. That’s what Cognition founder and CEO Scott Wu – maker of the AI coding agent Devin – suggests.
But even he admits that while Devin can work independently, he would currently rate his skills between a junior and mid-level developer, depending on the task. This is not a throw-it-and-forget-it solution.
SMU researchers suggest a more humane approach. Developers should know what tasks AI does and doesn’t do as deeply as they know their favorite coding languages. They need robust quality assurance systems designed for AI and are stuck with carefully reviewing AI work as if it were junior developers.
In the meantime, the researchers say (and Wu agrees), humans should still do the big-picture work, like software architecture and security design.
When you purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
