Dot reviewed my website copy and redid the whole site — vs my own agent
added
自己用心開發的 Agent 跟現成的 ChatGPT dot 差在哪? 一樣要求看我的 AI Agen 自學網找出文案的矛盾處和給建議,結果天差地別。 我自己開發的 AI Agent,雖然只用 35B 開源模型, 仍能有效找出文案衝突或不好懂的地方。 而 ChatGPT dot 號稱只用最強模型 GPT 6 Astra, 用了超久的時間,洋洋灑灑列一大堆要改的地方, 還重做了整個網站,把我本來花很多時間修改, 強調特色的文案,都換成平庸的 AI 文。 因為自己開發更聰明,超越 Astra 嗎? 不是。單純是更懂自己而已。 而自己開發,就是想讓 AI Agent 更懂自己。
First-hand comparative test (Mandarin): 陳盟升 (darks0603, 2.7K followers) gave the same task to his own self-built 35B open-source agent and to his ChatGPT Dot: review his AI-agent self-study website, find contradictions in the copy, and give suggestions. His own agent found the conflicts quickly; the Dot took a very long time, listed a huge set of changes, and redid the entire website — replacing his carefully hand-tuned, distinctive copy with generic AI prose. His verdict: the advantage of building yourself is that it understands you better, not that it's smarter than Astra.

