Inside Reflection: the new AI lab that's coming for DeepSeek
My full Sources Live interview with Reflection co-founder and CTO Ioannis Antonoglou in Davos.
Today I’m publishing my full interview with Reflection co-founder and CTO Ioannis Antonoglou, which took place two weeks ago in front of a live audience at Brunswick Home on the sidelines of the World Economic Forum in Davos, Switzerland.
Reflection is a new frontier AI lab that has raised over $2 billion to challenge the rise of open-source Chinese models like DeepSeek. Antonoglou tells me that the startup plans to release what it hopes will be “the most powerful open-weight model in the world” later this year.
Our conversation covers Reflection's mission to build open-weight AI models as an alternative to the closed approaches of competitors like OpenAI and Anthropic. Antonoglou also discusses his earlier work in reinforcement learning research at DeepMind, Reflection's business model around "sovereign AI" for enterprises and nations, and their close partnership with Nvidia.
Reflection isn’t a household name yet, but it’s generating a lot of buzz in the AI research community. If the startup manages to achieve its goal of being at the frontier of open-source AI this year, you’ll likely be hearing a lot more about it soon.
Stay tuned for more of my Sources Live interviews in Davos. And thanks to Disruptive for making this interview possible.
The following transcript has been lightly edited for length and clarity:
