A possibility is that they are betting that the current generation of LLM is converging, so they won't worry about the goalpost much. If it's true, then it won't be good news for OpenAI.