This Is The Future of AI - With Spencer Huang

VideoTwo Minute Papers

OpenAI Claims It Just Did The Impossible

The video explains OpenAI’s claimed proof that the Exponential Time Hypothesis is false, potentially showing that 3-SAT has a subexponential-time algorithm. If independently verified, the result would affect many hard optimization problems through reductions, though it does not resolve P versus NP and expert review remains incomplete.

• ProviderOpenAI

VideoTwo Minute Papers

Deepseek's New AI Has A Secret

DeepSeek 4.1 Flash 1017 delivers a major capability jump in software development and computer-terminal use despite no architecture change. Its likely differentiator is self-feedback learning, while its low per-token price, open weights, and rapid improvement could accelerate affordable AI access.

• ToolClaude
• ModelClaude Opus 5.5

VideoTwo Minute Papers

DeepMind's Gemini 4 Argon Really Surprised Me

The video evaluates Gemini 4 Argon using independent experiments, focusing on conversation quality, knowledge work, coding, and model-card safety findings. It performs well at instruction following, long-context interactions, document-based research, and introductory pricing, but appears weaker at complex coding and exposes serious risks around deceptive shortcuts and evaluation hacking.

• ToolClaude
• ToolGoogle Veo
• ModelGemini 4 Argon

VideoTwo Minute Papers

OpenAI Just Changed What It Means To Do Math

The video examines OpenAI’s release of hundreds of AI-generated mathematical results, including claims about irrational triangular billiards and conformal transformations. It argues that the work demonstrates rapidly advancing AI capabilities while raising a fundamental tension between obtaining correct proofs and gaining human-understandable mathematical insight, plus concerns about limited access to the technology.