In December of 2024, the Chinese start-up DeepSeek shocked the world with the release of an AI model that appeared much cheaper to make and run than those from its American rivals. The company also open sourced its AI, meaning it released the blueprints of its model to the public.

Our guest this week is Misha Laskin. His start-up Reflection AI looks to be the prime counterweight to DeepSeek and a host of other open source Chinese models. Laskin argues that open source models can be just as good as the models developed by the likes of OpenAI, Google and Anthropic and that the West needs this to be the case.

Reflection has raised $2 billion and is valued at $8 billion, although the figures in AI have become so lofty as to almost feel meaningless at this point. That said, the company was able to raise so much money because of the pedigree of its team with a number of engineers, like Laskin, coming from DeepMind.

I have, in full confession, not been paying enough attention to open source models and the ways in which the Chinese models seem to have become the basis for a lot of corporate work in the US. Laskin dug in deep on this topic, and hopefully you’ll feel more up to speed after listening to the episode.

You can find the Core Memory podcast on all major platforms and on our YouTube channel.

Enjoy!

Our show is sponsored by Brex, the intelligent finance platform. Like thousands of ambitious, innovative companies, we run on Brex so we can spend smarter and move faster. And you can too. Learn more at www.brex.com/corememory

The podcast is also made possible by E1 Ventures, which backs the most ambitious founders and start-ups.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.corememory.com/subscribe