Anthropic's Pre-training Lead and Midjourney's Medical Pivot
Nate B Jonesgo watch the original →
the gist
While OpenAI dominates headlines, Anthropic's focus on fresh pre-trained models and key talent acquisitions suggests a technical lead in recursive self-improvement, while Midjourney's move into affordable, preventative medical imaging represents a more significant real-world application of AI capital.
The Case for Anthropic's Technical Lead
While OpenAI captures public attention through post-training reasoning improvements and high-profile hires like Noam Shazeer, Anthropic is arguably better positioned due to its focus on large-scale pre-trained models. OpenAI has relied heavily on reasoning layers and post-training optimizations to iterate on its models, whereas Anthropic has maintained a more consistent cadence of releasing fresh, large-scale pre-trained models. This strategy is critical because pre-trained models inherently possess higher intelligence, and Anthropic's current lead in this area provides a superior foundation for recursive self-improvement. The recent hire of Nobel Prize winner John Jumper from Google further bolsters Anthropic's capacity to leverage these models for future breakthroughs.
Midjourney's Shift to Preventative Healthcare
Beyond the model-maker rivalry, the most significant development in the AI space is Midjourney's pivot into medical hardware. Leveraging its highly profitable, bootstrapped business model, Midjourney is developing a high-speed, affordable ultrasound device designed for preventative whole-body imaging. Unlike traditional medical imaging, which is reactive and expensive, this technology aims to enable population-level monitoring of cardiovascular health and early cancer detection. By operating outside the constraints of venture capital boards, Midjourney is applying its resources to a tangible, high-impact hardware problem that contrasts with the iterative model-race focus of the major labs.