As tech CEOs around the world advocate for AI-written code, Zoho Chief Scientist and co-founder Sridhar Vembu has offered a candid assessment of AI’s role in software development. In a post on X (formerly Twitter), Vembu detailed how his team scrutinised C++ code produced by Anthropic’s Claude Opus 4.5 model, highlighting both the strengths and limitations of current AI coding tools.
The session, which Vembu said extended for hours into the evening, focused on evaluating thousands of lines of AI-generated code.
In the post, Vembu noted that the model excels at “stitching together systems,” particularly in creating “glue code”. “I now have a much clearer understanding of what these models do well: they are able to stitch together systems well, taking data from one system, reshape it and pass it to another system. There is often a lot of such “glue code” in these systems and that is not very complicated but it is very tedious,” wrote Vembu.
Vembu points out strengths of AI in code generation
“AI models have ‘memorised’ all the open source [code] too and they are able to recall patterns from them (with some possibility of hallucination),” Vembu wrote. “They are also able to stitch various open source pieces together well,” he added.
The reviewed code, spanning several files with thousands of lines each, was mostly “straightforward glue code,” with only a small portion qualifying as complex. “I suspect that the AI-generated code tended to be needlessly verbose but I have to study it more to be sure,” he stated.
Humans still crucial to AI coding
Vembu emphasised the critical human element in AI coding. “Our senior engineer had guided (“orchestrated” is the right word) this process. When the AI was stuck he helped “unstuck” it,” he said. This shows that AI coding tools still need humans to help them guide through the complexities and come up with desired output.
“This was a very vital contribution and without his experienced guidance, the AI output would not be useful,” said Vembu.
“On the whole, I am both impressed and not super awed,” he concluded. “I believe we can do better.”
The post comes hours after Vembu himself shared an incident related to AI coding that left him shocked. In a previous post, Vembu has shared examples of dramatic productivity gains, such as a single engineer completing a year-long security tool project in one month using Claude Opus 4.5. Vembu had stated earlier that Zoho encourages engineer-led AI adoption, with mandatory human review of all generated code.
