Scaling AI Agents from Millions of Users: Technical Breakthrough or Market Bubble?

In July 2024, OpenAI announced that the total number of users for its agent products Codex and ChatGPT Work had surpassed 10 million, nearly doubling since the beginning of the month. This figure means that in just three weeks, the product line added about 5 million users, with a daily growth rate exceeding 3%. Compared to Anthropic's Claude Enterprise edition, which has publicly reported monthly active users of around 3 million, OpenAI's temporary lead in the agent space has created an order-of-magnitude difference. But as a researcher who has long focused on computer vision and multimodal systems, what I care more about is: behind this growth, is there a true breakthrough in the technology paradigm, or is it just a short-term bonus during the market education phase?