What I Read This Week…

OpenAI releases o1 and o1 pro mode, the FDA's 'generally recognized as safe' category creates oversight gaps, and scientists find that stimulating the hypothalamus may help spinal cord injury patients

Watch All-In E206 Read our Deep Dive: A Primer on Cryptocurrencies Caught My Eye… OpenAI has released o1 and o1 pro mode, which both use the base o1 model's chain-of-thought reasoning to process responses. OpenAI has reported that o1 operates 50% faster than o1-preview, reduces major errors by 34% compared to o1-preview, and can analyze both text and images. While o1 pro mode takes longer to process responses than o1, it uses more compute power and can provide more reliable answers than o1 for highly demanding technical queries in fields like mathematics, chemistry, and programming. As a…

related reading

• Reverse engineering OpenAI’s o1interconnects.ai
• Gap Mapgap-map.org
• o1 System Cardassets.ctfassets.net
• Learning to reason with LLMs | OpenAIopenai.com
• Late Takes on OpenAI o1alexirpan.com
• o3, Oh Mythezvi.substack.com
• AI #184: Post Post Mortemthezvi.substack.com
• The o1 System Card Is Not About o1 — LessWronglesswrong.com
• o1: A Technical Primer — LessWronglesswrong.com
• How AI Is Learning to Think in Secretnickandresen.substack.com
• Two different tricks for fast LLM inferenceseangoedecke.com
• Security incident disclosure — July 2026huggingface.co