AI's vibe-coding era: How the shift to apps changed the race
Value within the artificial intelligence industry is slowly shifting, from the companies developing models to the apps building on top of them.
Early in the AI race, critics viewed apps like Perplexity, Replit, Sesame and Abridge as second-rate middlemen, slapping an interface on someone else's technology. They were disparagingly known as AI wrappers: companies with entire apps or businesses wrapped around existing models. Companies like OpenAI, Google, Meta and Anthropic developed their own models.
The arrival of ultra-efficient models and increasing model commoditization accelerated the shift.
"There was an impression that the only way to compete in AI would be to raise hundreds of millions of dollars to pre-train these web-scale models that could solve every problem underneath the sun, and that was the only game in town for AI," said Shiv Rao, founder and CEO of the healthcare AI startup Abridge. "Very quickly, people figured out that actually, value moves up the stack."
Megacaps like Microsoft poured billions into the first stage of the AI arms race, focusing on the infrastructure and model layer. But models are now increasingly looking commoditized, narrowing the advantage that any model-builder had. While they focused on delivering raw capability and intelligence, app companies looked at real-world uses and solutions.