Devin Won't Save You: Why Devs Are Now Too Addicted to Code Without AI
Turns out, the future of coding is just us being completely unable to function without our AI crutches. Researchers at METR found that developers have essentially ghosted manual work, even though the tools are actually making us slower.
In early 2026, METR discovered that the modern coder has hit a wall of dependency: they simply refuse to work without AI assistance, even in controlled experiments. This isn't because the code is better—in fact, AI tools often slow down development by forcing engineers to spend endless hours fixing machine-generated errors and managing prompts.
The industry trend of 'token-maxing'—treating total AI usage as a metric of productivity—hit a massive reality check. Amazon quietly shuttered its Kirorank leaderboard after realizing employees were just spamming AI agents to fluff their stats, while Uber burned through its entire annual AI budget in four months with zero measurable uptick in output. As James Shore pointed out, companies are effectively trading temporary speed for a permanent, high-cost maintenance debt.
Data from CodeRabbit suggests that AI generates 1.7 times more bugs than humans, and Entelligence AI founder Aishwarya Sankar estimates nearly half of all compute tokens are wasted just cleaning up this robotic mess. Even Scott Wu, the architect behind Devin, concedes that his creation functions more like a shaky junior developer than a magic 'set-it-and-forget-it' oracle. The Singapore Management University research team suggests that instead of treating AI like a senior engineer, we should start auditing its output with the same skepticism we reserve for interns.
The industry has reached a state where the tools designed to automate labor have become the primary source of labor themselves. It is a brilliant irony that tech companies are now paying exorbitant cloud fees to manage the broken code that their own expensive AI models created. The real cost isn't the subscription fee; it's the total atrophy of the human ability to solve a logic puzzle without a chatbot holding the hand.
Source: TechCrunch
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.