The discussion over whether artificial intelligence truly improves software developer productivity is becoming more intense as tech companies continue pouring money into AI-based coding assistants and tools. Zoho chief Sridhar Vembu recently weighed in on the topic through a post on X, where he spoke about what he called the software industry’s “developer productivity paradox.”
Citing a discussion from Hacker News about AI-generated code at scale, Vembu argued that while large language models can draw from an enormous pool of information, they still do not genuinely understand the real problem they are trying to solve.
Sridhar Vembu connected the debate to remarks made by François Chollet, a French AI researcher and software engineer best known for creating the Keras deep learning framework, who pointed out that although developers are writing far more code with the help of AI tools, the actual gains in productivity appear far smaller than expected.
Chollet argued that programmers are increasingly producing large volumes of code for relatively minor tasks instead of delivering substantial improvements or real-world impact. He also noted that AI-generated code can sometimes create fresh problems, forcing developers to spend extra time fixing errors later.
Referring to a Hacker News discussion on the subject, Vembu said the comments perfectly captured what he described as the software industry’s current “developer productivity paradox.”
“It is capable of applying [an] incredible amount of knowledge while having virtually no real understanding of the problem,” Vembu wrote, referring to AI-generated code.
Companies are heavily investing in AI:
Sridhar Vembu further said that businesses around the world are pouring hundreds of billions of dollars into AI, believing these tools could eventually boost software engineering productivity by ten times or even more. Despite those expectations, he noted that the improvements seen until now have been relatively modest.
The conversation has since expanded into a broader industry debate over whether AI-powered coding assistants are genuinely improving software quality or merely increasing the volume of code developers produce.
Social media reacts:
Many participants in the discussion argued that the toughest parts of software engineering still involve areas such as planning, architecture, problem-solving, and understanding complex systems — tasks where human judgment remains critical. Others, however, said AI tools are proving useful in speeding up testing, experimentation, and repeated development cycles, helping teams build and refine applications more efficiently and at lower cost.