Teortaxes Weighs OpenAI Reasoning RL Against Anthropic Lead

@teortaxesTex, the pseudonymous X account known as DeepSeek's most vocal online supporter, posted a clarification on model development paths. The account states the comment is not a prediction that OpenAI will win overall. It notes Anthropic gained a temporary edge through faster pretraining and agentic coding work. The post adds that OpenAI appears to hold a stronger method for reasoning RL. It also points to Chinese results as evidence that scale may be overrated for capabilities tied to recursive self-improvement.