Will the labs devour the apps that run on their models?
|SAN FRANCISCO|5 min read
IN THE USUALLY gossipy world of Silicon Valley, something strange is happening. It is hard to find a generative artificial-intelligence entrepreneur with a bad word to say about anyone. This may be an age thing. Many of those launching AI startups were born after sci-fi dystopias like “The Matrix” (né 1999) and are young enough to still believe AI will be a force for good for everyone. To them, even the word “frenemy” sounds too red in tooth and claw.
Bonhomie aside, though, a competitive dynamic is emerging that will only grow fiercer. Take AI startups like Anysphere, whose Cursor app helps software developers write code, Harvey, which provides AI to law firms, and OpenEvidence, which does the same for doctors (and on October 20th raised $200m at a valuation of $6bn). These apps are growing rapidly by making use of large language models (LLMs) provided by the likes of OpenAI, maker of ChatGPT, and Anthropic, maker of Claude. They pay to use the models, but unlike the AI labs, they do not burn through billions in a quest for superintelligence. That makes their path to profitability easier.
This article appeared in the Business section of the print edition under the headline “The Cronos syndrome”
From the October 25th 2025 edition
Discover stories from this section and more in the list of contents