Developers won't code without AI now, and the bill is coming due
Coders increasingly refuse to work without AI assistants, yet the speed comes with a hidden maintenance tab. For India's IT services giants, that tab could be enormous.
The News
Software developers have grown so attached to their AI assistants that many now refuse to work without them, even as fresh evidence suggests the tools are quietly inflating the cost of the code they help write. TechCrunch reported on 29 May 2026 that the research lab METR discovered something striking back in February 2026: developers were declining to take part in its studies if it meant switching the AI off. In the lab's words, they simply do not wish to work without it.
The productivity story those developers tell themselves is generous. A METR survey in May 2026 found coders believe the tools have made them roughly twice as valuable to their employers. The harder numbers point the other way. An analysis from CodeRabbit found AI-written code produced 1.7 times more problems than code written by humans. Aiswarya Sankar, founder of Entelligence AI, said some companies are now spending 44% of their tokens just fixing bugs the models generated.
The budget signals are blunt. Uber reportedly burned through its entire 2026 AI allocation in the first four months of the year, and Amazon quietly shut down an internal leaderboard, Kirorank, that had been tracking engineers' token consumption.
Why It Matters
The uncomfortable finding is that faster is not the same as better. A Singapore Management University report in April 2026 warned that AI-written code can introduce long-term maintenance costs into real software projects, echoing an earlier 2025 METR study that found the tools actually slowed experienced developers down even as they shipped code more quickly. Programmer and author James Shore put the trade plainly: write code twice as quickly, and you had better hope you have halved your maintenance bill too.
This is the second act of every productivity boom in software. When offshore outsourcing scaled in the 2000s, raw delivery speed masked a quality reckoning that arrived years later in the form of brittle, hard-to-maintain systems. The AI coding wave, led by tools such as Cognition's Devin agent under chief executive Scott Wu, is compressing that same cycle into months rather than years. The speed is real. So is the debt it leaves behind.
Indian Angle
Nowhere does this matter more than India, whose IT services sector is built on the economics of writing and, crucially, maintaining other companies' code. Firms such as TCS, Infosys, Wipro and HCLTech earn a substantial slice of revenue from application maintenance and support, often on multi-year contracts. If AI-assisted code genuinely carries a heavier maintenance load, that is paradoxically more billable work, but only if clients keep paying. On fixed-bid deals, the same buggy output becomes a margin drain the vendor must absorb.
India also has the largest concentration of developers on the planet and a fast-growing base of Global Capability Centres run by global firms. A generation of Indian engineers is now learning the craft with an AI co-pilot always switched on, which makes the dependency METR observed a structural workforce question. NASSCOM's repeated emphasis on upskilling looks less like a slogan and more like a hedge.
There is a regulatory shadow too. As Indian banks, insurers and public-sector bodies push more software into production, RBI and CERT-In scrutiny of code quality and security will sharpen. Code that ships fast but breaks later is exactly the kind of risk those bodies exist to catch.
FAQ
What did METR actually find?
METR, an AI research lab, found in February 2026 that developers were unwilling to join its studies if AI tools were removed. A separate METR survey in May 2026 showed coders believe AI makes them about twice as valuable, a perception its own 2025 work had already questioned.
Is AI code really buggier?
The evidence is mounting. CodeRabbit's analysis found AI-written code produced 1.7 times more problems than human code, and Entelligence AI reported some firms spend 44% of their tokens fixing AI-created bugs. Speed appears to come at the cost of quality.
Why should Indian IT firms care?
India's services giants earn heavily from maintaining software. Higher maintenance loads mean more work on time-and-materials contracts but eroded margins on fixed-price deals, making code quality a direct profitability issue rather than an abstract concern.
Where can I read the original coverage?
The reporting is by Julie Bort for TechCrunch, published on 29 May 2026, and is linked in full in the attribution below.
This story was reported by TechCrunch. Read the full original coverage at TechCrunch.