Zoho CEO questions scalability of vibe-coded tools after Garry Tan predicts disruption of traditional SaaS

Sridhar Vembu

Zoho CEO Sridhar Vembu has pushed back against Y Combinator President Garry Tan’s prediction that “vibe-coded” applications — software assembled rapidly using AI and lightweight developer tools — could outpace traditional SaaS offerings. Tan had argued that companies may soon use platforms like Replit, Emergent Labs and Taskade to build their own custom workflows instead of paying for bundled SaaS subscriptions.

Responding on X, Vembu questioned why Zoho would be among the first to be disrupted if the shift were already underway. He pointed to Zoho’s customer base growing at more than 50 percent, noting that businesses continue to choose established software ecosystems rather than depending on weekend-built internal tools.

He also challenged the absence of credible vibe-coded alternatives in essential categories such as email, spreadsheets, accounting and messaging. According to Vembu, these require deep engineering work, robust data handling and decades of accumulated product stability — qualities that rapid AI-led development currently cannot match.

‘Vibe coding accumulates tech debt until collapse,’ says Vembu

In his detailed post, Vembu said the shortcomings of vibe-coded tools stem from a lack of guarantees around compliance, privacy, security and long-term maintenance. He cautioned that applications built this way tend to amass technical debt faster than organisations can manage, eventually hitting a point where they break under real-world demands.

“Without those guarantees, vibe coding just piles up tech debt faster and faster until the whole thing collapses,” he wrote, adding that enterprises need systems that function reliably under audits, uptime requirements and scaled usage — not prototypes assembled over a weekend.

Vembu also took aim at Tan’s remark about startups offloading tech debt onto acquirers, suggesting that such thinking overlooks the operational realities of enterprise software.

Several users on X echoed Vembu’s view. Some noted that while vibe coding works for MVPs, running payroll, handling sensitive customer data or meeting compliance audits demands far more rigour. Others shared examples of teams abandoning vibe-coded internal tools after data issues, script failures and scaling bottlenecks.

Vembu concluded by outlining his own R&D direction: combining AI with compiler technology to boost developer productivity while maintaining the reliability enterprises expect. He said this approach offers a more durable path than relying solely on AI-generated shortcuts.

The exchange highlights a broader moment of tension within the software industry as AI-driven development tools evolve, raising questions about how far they can scale without foundational engineering discipline.