#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI

with Sebastian Raschka & Nathan Lambert

1 Feb 2026 18 min read 2h 5m

TL;DR

The AI landscape in early 2026 is defined by intense competition between US closed-weight models (OpenAI, Google, Anthropic) and Chinese open-weight alternatives (DeepSeek, Qwen, MiniMax), with no clear winner emerging. While US models currently maintain superior intelligence and speed advantages, Chinese open models are proliferating rapidly and gaining influence through accessibility, forcing differentiation through infrastructure, organizational culture, and specialized capabilities like coding and long-context retrieval.

Key Moments

Sebastian Raschka

“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. That is mainly because researchers are frequently changing jobs and labs. They rotate. I don't think there will be a clear winner in terms of technology access. However, I do think there will be, The differentiating factor will be budget and hardware constraints.”

Responding to the question of who is winning at the international level between China and the US

▶ 3:16

Nathan Lambert

“Anthropic seems to at least be presenting as the least chaotic. It's a bit of an advantage, if they can keep doing that for a while. But on the other side of things, there's a lot of ominous technology from China where there's way more labs than DeepSeek.”

Discussing how organizational culture and execution differentiate players despite fluid idea space

▶ 5:04

Nathan Lambert

“I think Gemini will continue to make progress on ChatGPT. I think Google's scale, when both of these are operating at such extreme scales—and Google has the ability to separate research and product a bit better, whereas you hear so much about OpenAI being chaotic operationally and chasing the high-impact thing, which is a very startup culture.”

Making a prediction about who will win in 2026 on the consumer chatbot front

▶ 11:52

Sebastian Raschka

“Building an LLM from scratch is a lot of fun. It's also a lot to learn. And like you said, it's probably the best way to learn how something really works, 'cause you can look at figures, but figures can have mistakes. You can look at concepts and explanations, but you might misunderstand them. But if there is code, and the code works, you know it's correct.”

Explaining why building systems from scratch is the best pedagogical approach for understanding AI

▶ 24:27

Sebastian Raschka

“One of the most common complaints about LLMs are, for example, hallucinations, right? And so, in my opinion, one of the best ways to solve hallucinations is to not try to always remember information or make things up. For math, why not use a calculator app or Python? If I ask the LLM, "Who won the soccer World Cup in 1998?" instead of just trying to memorize, it could go do a search.”

Discussing why gpt-oss-120b's tool-use capability represents a paradigm shift for addressing hallucinations

▶ 32:36

About the guests

›

Sebastian Raschka is a machine learning researcher and author of "Build a Large Language Model from Scratch" and "Build a Reasoning Model from Scratch." Nathan Lambert is the post-training lead at the Allen Institute for AI and author of the definitive book on Reinforcement Learning from Human Feedback. Both are widely respected researchers, engineers, educators, and influential voices on AI development and open-weight models.

Takeaways

Infrastructure & resources trump proprietary tech The fastest innovation spreads across labs as researchers move jobs, making technical ideas non-proprietary. Winners will be determined by GPU access, data center scale, and training budgets rather than algorithm secrecy. Google's vertical integration (TPUs, custom chips) and OpenAI's research velocity give them structural advantages unrelated to secret sauce.

Chinese open models reshape competitive dynamics DeepSeek's 2025 breakthrough catalyzed a wave of open-weight releases from Qwen, MiniMax, and Kimi that now threaten US API monetization. While US closed-weight models maintain intelligence/speed advantages today, Chinese firms are building influence internationally and challenging the assumption that AI users will pay subscription fees—a shift particularly significant outside North America.

Tool use unlocks the next LLM reliability frontier gpt-oss-120b and similar models trained with web search and code execution capabilities represent a paradigm shift beyond pure memorization. Rather than solving hallucinations through scaling, tool-augmented LLMs can ground answers in real-time data and computation, fundamentally changing how users interact with models and what reliability means.