I blindly guess that GPT 5 will be a big mess tonight. I have already lost trust in Aotuman🥲 Personally, I feel that GPT is becoming more and more mediocre. In summary, there are probably the following reasons:
- Too many core members have left, early executives have left one after another, and meta has been poaching people like crazy recently.
- It swings between commercialization and open source. It is not focused and is far less focused than anthropopic. In the past six months, it has even been overtaken by Gemini.
- Aotuman should be a feminine person with strong business talent, but he may not necessarily understand researchers and engineers, which may cause many problems, especially for technology startups. In terms of 4 large model routes, GPT does not seem to place coding capabilities in an important position and is too obsessed with AGI. 5 Application-side product packaging has always been a weakness of GPT, such as GPTs, the dazzling naming method of large models, o3 is very strong but too expensive pricing has discouraged many people. 6 GPT has almost 0 derivatives (spanning Sora and GPTs). During the same period, in addition to the large model itself, Anthropic launched around coding capabilities: MCP, ClaudeCode, Gemini, using the Google ecosystem, launched notebooklm, veo3, flow... 7 The traffic entrance is single, only ChatGPT, always resting on its laurels, and the API has no price advantage.
- There is a lack of a great AI product manager. My suggestion is to spend some money to recruit a product boss. This area is really too weak. For example, those well-known AI agent product managers. 9 The large model route does not seem to be closely integrated with the agent. This is not in line with the trend. Large models can be admired from afar, but we must also consider the large amount of implicit demand after the agent in front of us explodes on the application side. 10 Maybe Auguman is too arbitrary. Recently, I have seen that GPT does not share as much useful content as other companies.