Dot caught and fixed its own document mistake, unprompted

@hiroki1m

added Oct 2, 2026

Dotがすごい。 書類を整理して先方に送る作業をDotに任せたあと、確認のために「参考資料を開いて」と指示した。 すると、こっちは何も言ってないのに、 「さっき整理した書類に間違いがありました」と自分で気づいて修正を始めた。 これ、簡単に言うと 「上司の行動を見て、自分のミスを察して、自らリカバリーに動いた」ってこと。 従来の生成AIとは、ちょっと違う動き。 進化に感動した。

First-hand result from Hiroki Maejima (Giving Inc., Japan): he entrusted his Dot with organizing documents for sending to a client, then asked it to open reference materials for verification. Without being asked, the Dot announced it had found an error in the document it had just organized and started fixing it on its own. He frames it as the agent watching its 'boss's behavior, sensing its own mistake, and moving to recover by itself — behavior he says differs from conventional generative AI.