Interesting heatmap. The AI war isn’t just about models anymore, it’s increasingly about tight vertical integration (with Google and Apple clearly leading the way). Esp. with the X acquisition, I think the diagram could use one more layer: proprietary data-it’s becoming a major strategic moat.

Not sure about how Studio Ghibli would respond to this buzz now-should they be happy (seeing it as flattering free marketing), or be rightfully upset that their creative work is being replicated and spread across the internet with zero control or dollar paid? A big question ahead for the creatives

Yes, #DeepResearch may replace researchers in the future, but even today’s research work, verifying second-hand info is crucial. How can end users fully trust AI outputs in a world full of AI-generated info? As the saying goes, the journey (process) is just as important as the destination (answers)

Breakfast on #SandHillRoad with entrepreneurs and tech execs discussing #data, scaling, #RAG, speed-to-market, and #enterprise AI adoption. Interesting discovery: while big techs seem to have endless resources, they too face internal constraints and tough #prioritization challenges.

Tried Deepseek, the Chinese AI model trained for just $5M (11x less compute cost). Lightning-fast, at least on par with #Groq or #SambaNova Cloud imv. Interestingly their API directly supports #OpenAI SDK, making it seamless for devs to switch to Deepseek - a smart way to challenge OAI’s dominance

Jensen unveiled #WFM (world foundation model) at #CES, but I’m skeptical if it’s truly #physicalAI or just a marketing tactic. It’s combining #transformer-based diffusion, autoregressive models, RL/policies, and millions of video hours for training. Maybe call it “plausible physics model” instead?