The State of AI in 2026: LLMs, Geopolitics, and the Future of Human-AI Collaboration

This episode provides an expert analysis of the artificial intelligence landscape in early 2026, focusing on the intense competition between US and Chinese companies, the evolution of Large Language Models (LLMs), and the critical role of open-weight models. The discussion highlights the 'DeepSeek moment' of January 2025, when the Chinese company DeepSeek released a near-state-of-the-art open-weight model with significantly less compute, catalyzing an acceleration in AI development and competition. While the US currently leads in closed-weight frontier models like OpenAI's ChatGPT and Google's Gemini, Chinese companies are rapidly gaining ground, particularly in the open-weight space, driven by strategic incentives to build international influence and overcome API subscription security concerns from Western companies.\n\nThe conversation delves into the nuances of model differentiation, user preferences, and the underlying infrastructure. It's noted that technological ideas are highly fluid due to researcher mobility, making budget and hardware constraints the primary differentiating factors rather than proprietary algorithms. The hosts and guests discuss the hype cycles around models like Anthropic's Claude Opus 4.5 and Google's Gemini 3, emphasizing that while specific models might capture 'mindshare' for a time, the overall landscape is characterized by rapid leapfrogging. A key distinction is drawn between intelligence and speed in LLMs, with users often preferring faster, 'non-thinking' models for quick daily tasks and more thorough, 'thinking' models for complex problem-solving or deep analysis.\n\nPractical insights emerge regarding the optimal use of LLMs. Users often employ multiple models for different purposes, such as Gemini for quick information, Claude Opus for coding and philosophical discussions, and specialized tools like Grok 4 Heavy for debugging. The concept of 'muscle memory' plays a significant role in user adoption, with established platforms like ChatGPT benefiting from long-term usage habits. The discussion also touches upon the transformative potential of LLMs in programming, advocating for 'programming with English' as a higher-level design approach, and in learning, where LLMs can enrich the reading experience by providing context and accelerating 'aha moments.'\n\nLooking ahead to 2026, predictions suggest continued progress for Gemini against ChatGPT, leveraging Google's immense scale and ability to separate research from product development, as well as its advantage in proprietary hardware like TPUs. Anthropic is expected to maintain success in the enterprise and software sectors, particularly with its code-focused offerings. The proliferation of open-weight models from China is anticipated to continue, fostering creativity and potentially leading to consolidation in the long term. The overarching implication is a future where AI technology is increasingly accessible, but competitive advantage will hinge on resource allocation, strategic business models, and the ability to optimize for diverse user needs across a rapidly evolving technological frontier." "concepts": [ "DeepSeek moment

I don't think nowadays, in 2026, that there will be any company that has access to technology that no other company has access to.

The differentiating factor will be budget and hardware constraints. I don't think the ideas will be proprietary, but rather the resources needed to implement them.

culturally Anthropic is known for betting very hard on code, which is the Claude Code thing, is working out for them right now.

DeepSeek kicked off a movement within China, I say kind of similar to how ChatGPT kicked off a movement in the US where everything had a chatbot.

a lot of top US tech companies and other IT companies won't pay for an API subscription to Chinese companies for security concerns.

I think the momentum, if you look at 2025, was on Gemini's side, but they were starting from such a low point.

Google has a just kind of historical advantage there. And if there's going to be a new paradigm, it's most likely to come from OpenAI where their research division again and again has shown this ability to land a new research idea or a product.

I think it's genuinely more fun to program with an LLM. And I think it's genuinely more fun to read with an LLM.

But if there is code, and the code works, you know it's correct. I mean, there's no misunderstanding. It's precise. Otherwise, it wouldn't work.

the US models are currently better, and we use them. I tried these other open models, and I'm like, 'Fun, but I don't go back.'