OpenAI CEO Sam Altman announced that the company's internal AI coding models have advanced to rank among the top 50 competitive programmers globally, with a Codeforces rating of 3045, marking a rapid progression in their capabilities. The AI models have climbed from the 9,800th position with model o1 in October 2023 to 175th with o3 in December 2023, which had a rating of 2727 on Codeforces, and now stand at 50th as of February 2025. Altman expressed confidence that the AI will achieve the number one ranking by the end of 2025, potentially surpassing the abilities of the programmers who created it. Altman outlined plans to further enhance these models through pretraining, reinforcement learning, and scaling using the Stargate infrastructure, which offers 100 times more computational power. This initiative aims to transition from the current internal GPT-4.5 to GPT-5.5, potentially leading to new discoveries in physics and biology. In a recent blog post titled "Three Observations," Altman discussed the unprecedented rate at which AI is scaling and its impending impact on society and the economy. He noted that the cost to use AI decreases tenfold annually, leading to significantly increased usage. Altman emphasized that Artificial General Intelligence (AGI) is on the horizon and will profoundly affect capital and labor. He added that current AI systems excel in STEM fields and are set to become equally proficient in other areas within the next six to twelve months.