Sorry, Garry! YC’s Vibe-Coding Startups ‘Aren’t Replacing Zoho’ Anytime Soon

A war of words between Zoho’s Sridhar Vembu and YC’s Garry Tan over vibe coding has opened a broader debate about its practicality for large-scale software development.  

The Y Combinator president claimed that Zoho-type SaaS is on borrowed time and can soon be replaced by custom tools built by non-technical teams on Replit, Emergent Labs or Taskade.

“Why pay $30/seat/month for over-bundled SaaS when soon even nontech ops people can vibe-code a custom solution in a weekend?” Tan quipped. Vembu did not hold back and challenged Tan that he would “outlast and outshine” his vibe coding companies. 

“If our business were the first to be competed away by vibe-coded apps, why are we seeing such rapid customer growth (exceeding 50%) right now?” he wrote. Vembu added that essential software categories remain untouched by vibe coding tools. “Why don’t we see vibe-coded email or spreadsheet, accounting app, or messaging apps yet?”

He argued that security, privacy and compliance are central to any long-term productivity gains. “Without those guarantees, vibe coding just piles up tech debt faster and faster until the whole thing collapses,” the Indian SaaS company founder added. 

Some users on X also found a humorous side to the debate. “I will vibe code a cheaper Replit, Emergent Labs and Taskade replacement,” a user posted on X.

There have been multiple recent cases in which vibe coding tools have malfunctioned. A photographer using Google’s Antigravity tool reported on Reddit that the AI had erased his entire Windows D: drive after automatically generating and running a command.

The scepticism isn’t limited to Vembu. Abhinav Girdhar, founder and CEO of Appy Pie, told AIM that vibe coding remains limited in its current form. “At the moment, vibe coding works best for prototypes, workflow automations, and well-bounded features. For enterprise-grade systems with high scalability or regulatory needs, vibe coding can accelerate development but not fully own it,” he said.

While it can speed up development, Girdhar said that organisations still need engineers to define architecture, maintain quality, and ensure long-term stability. According to him, the near future will likely see hybrid models where AI handles repetitive logic and human teams focus on system design, integrations, and governance.

Similarly, Deepak Dhanak, co-founder of Rocket.new, an AI-native platform, told AIM that he believes the debate is fundamentally misunderstood. He said that the conversation is being framed as a binary where vibe coding either replaces SaaS or gets dismissed as a toy. In his view, this framing is simply wrong.

Vibe Coding Is a New Layer — Not a Replacement

Dhanak said that vibe coding does not change the fundamentals of software development. He says the underlying work, such as infra, databases, testing, deployment, and scaling, still remains. The only change is that AI now brings all these layers into a single interface instead of scattering them across multiple tools.

The real issue, Dhanak argued, isn’t vibe coding and prompting itself but the uneven quality of platforms in the space. Some AI coding tools are so lightweight that he calls them toys, useful only for quick prototypes and nowhere near production-grade.

He stressed that the limitations of vibe coding become clearer in the context of backend systems. 

Dhanak says the backend remains the real engine of any business — the place where core data lives, workflows run, governance happens, and payments or transactions are processed. “It is definitely going to take some time for the current technology to reach a point where you can vibe code your backend,” he said. 

Zoho is Not Going Anywhere

Experts believe that the SaaS company could simply add vibe coding capabilities on top of its existing services, allowing customers to customise and personalise their tools as needed.

“Vibe coding lowers the barrier for new entrants, which could pressure established vendors. But for companies offering tightly integrated ecosystems—like Zoho—this shift can actually be an advantage. They can embed AI-driven customisation directly into their platforms, making it easier for customers to tailor apps without external developers,” Girdhar said.

At the same time, Dhanak said that the future of SaaS will look very different as customers demand more control and personalisation. Instead of buying full-fledged SaaS applications, enterprises may increasingly rely on backend APIs and build their own front-end layers using vibe coding tools. 

He explained that the ideal setup is one in which Zoho provides the backend via APIs, while customers build their own front-end applications on top of it. This kind of model, he believes, will push SaaS toward API-first consumption rather than traditional app-based subscriptions.

Notably, Zoho recently announced a series of AI enhancements to its low-code application development platform, Zoho Creator. The tool enables users to build and refine applications using written prompts, process diagrams, or business specification documents. 

The feature is powered by Zoho’s AI engine, Zia, which aims to reduce the time required to launch applications and make the development process more accessible.

Bharath Kumar, head of marketing and customer experience at Zoho Creator, in a previous interaction with AIM, said the company wants to empower developers rather than threaten their work. “What AI is helping with boils down to just two fundamental things—it is either helping you do things faster, or it is helping you do things at a better quality.” 

What about Technical Debt?

On the much-discussed issue of technical debt, Dhanak acknowledged the concern but argued that the problem largely stems from the use of undisciplined or generic tools. He warns that many new AI coding platforms lack a clear architectural philosophy, inevitably leading to mounting debt.

Girdhar shared a similar view, noting that technical debt becomes a real issue when AI-generated code initially works but becomes difficult to modify or extend. “Long-term sustainability will depend on two things: whether AI systems become better at producing standardised, interpretable structures, and whether organisations invest in governance around AI-generated code,” he said.