Modelence raises $3 million to power its vibe coding stack

As AI tools democratize software engineering, a new generation of users has emerged eager to build their own apps. But as much as an LLM speeds up the coding process, old questions about hosting, security, and general DevOps issues still remain.

There is clearly a business opportunity in solving this problem, but the rapid changes in the system make it difficult to seize it.

One of the more interesting answers comes from Modelence, a Y Combinator startup from this summer’s batch that announced Wednesday that it had raised $3 million in a seed round. Y Combinator led the round, with participation from Rebel Fund, Acacia Venture Capital Partners, Formosa VC, and Vocal Ventures.

Moderence isn’t the only company targeting this area. Both giants like Google and Amazon and small startups like Shuttle are trying to solve infrastructure problems.

California-based Modelence stands out for how it diagnoses problems. For CEO Aram Shatakhtsyan, the problem isn’t individual services. It’s the connection between them.

“You don’t want to ask artificial intelligence to build authentication, set up databases, and connect them together, because that’s very likely to break,” Shatakhtsyan told TechCrunch in a recent interview.

This is an interesting diagnosis that explains why so many world-class service providers contribute to such an unstable system.

“Vercel covers most of the front end, Supabase covers the database and the layers above it, but you still need to piece together the rest,” he said, explaining the various products software engineers use today. “In the best case, you get two cloud systems.” The bottom line is there’s a lot of room to make mistakes.

Modelence’s approach is to provide something like an all-in-one service. The company’s framework, or toolkit, runs on TypeScript, and the company handles authentication, database, hosting, LLM observability tools, and even its own Lovable-style app builder to eliminate extra friction.

It’s an interesting idea, so I’m looking forward to seeing if it can attract users. But with the landscape of code-adjacent tools changing so rapidly, it’s very difficult to keep up.