China Morning Missive

For those following the pitched battle over AI, this update out of Meta is so very telling. After first deciding to prioritize an open architecture, it is now looking increasingly likely that the new team will shift to a closed model.

Quote from the linked TechCrunch article sums it all up.

“Top members of Meta’s new Superintelligence Lab discussed pivoting away from the company’s powerful open source AI model, Behemoth, and instead developing a closed model, reports the New York Times.”

The reason for bringing this shift by Meta up is given other developments in the AI space. There’s been yet another open source model released out of China over the weekend – Kimi K2 – released by a group called Moonshot (perfect name!). From the commentary, it would appear as though Kimi K2 has leaned heavily on the DeepSeek model. This! This! And This! The key feature of open source, as we here all know, is that you take what it is that others have created and then make the models perform better.

Competition in it rawest form. Demand and usage will only gravitate to the model that is iterative in the overall development.

American AI platforms are seeking a walled garden and I am old enough to remember how that turned out of America Online (AOL)

Chinese AI platforms see the threat posed by the American closed models and simply made the decision to commoditize the entire industry. Go big. Build strong. Make the American AI models obsolete in short order.

There was evidence of this approach with the launch of DeepSeek. Today, however, the writing is clearly on the wall. So long as the American AI players work within corporate silos, the Chinese are destined to win the global AI race. That fact is now far more evident than ever.

https://techcrunch.com/2025/07/14/meta-built-its-ai-reputation-on-openness-that-may-be-changing/