Owning your own intelligence: why open-weight models are the next battleground

Key points

Key takeaways from Source interview with BaseTen co-founder and CEO Tuhin Srivastava on The Sachin and Adam Show (October 2026):

AI research is concentrated. Most frontier research is done by a very small set of labs focused on a specific slice of the AI stack, leaving the wider ecosystem dependent on them for almost all intelligence capacity.

Open weights are getting very good. As open-weight models improve, more companies will start to own their own intelligence rather than renting it from a handful of providers.

Owning intelligence is the next frontier. Srivastava argues the growth area will be frontier work that figures out how to use a company's own data, workflows and user signals to make open-weight models perform at the frontier.

The internal compounding advantage. That work turns the compounding advantage companies accumulate inside — proprietary data and operating knowledge — into model performance, which customers say is a massive shift.

A future beyond two or three labs. The vision is a step towards a world where not every company is dependent on two or three labs for all of its intelligence, spreading capability across the economy.

Deployment infrastructure as enabler. As the CEO of an inference platform, Srivastava frames the platform's role as making it practical for companies to serve and tune open-weight models on their own terms.

Read more: The Sachin and Adam Show (YouTube)