While I agree with that to some degree, I believe it's not that they can't make better models, it's just the improvements aren't that spectacular, and OpenAI ne

While I agree with that to some degree, I believe it's not that they can't make better models, it's just the improvements aren't that spectacular, and OpenAI needs to be spectacular, they can't afford to be boring tech like Gemini. (I tested Claude 3.5 Sonnet on my custom task set recently and it beat ChatGPT-4 in several tiny but important details which means there is definitely a room for improvement for OpenAI)