Flowith is like an intern who is not capable enough and not smart enough, but always wants to do more, maybe he can please his boss. A task copy can only achieve 60 points, and before it can be changed, it starts to be busy generating pictures and soundtracks, and it misses the point at all. Genspark is like a small town as a problemalist, relying on pre-filled content to brush up on the sense of existence to get a good impression, and the mock questions are okay, but the on-the-spot response is almost meaningless. Eighty points have no hard injuries, and there are no memory points. But Manus, I'm really going to praise you. Not only can you polish every task, but you can also remember what I said (the knowledge base is so powerful that it will ask you "if you want to set it as a preference" every time you have additional needs). You can even say "Try it next time...", just like you would with an intern, and add it to the knowledge base with one click. Does anyone understand this feeling of blurring the boundaries between AI and humans? A product manager who feels good must be a good employee/colleague when he is young. Although the meaning of "part-time work" is often devalued in this day and age, it is only when we have seriously thought about how to work together in a group that we can understand how valuable the most basic human cooperative virtues are such as "aligning up and down, left and right" and "understanding the intention of the task". (It's really not Manus, it's just an annual user who loves assessments)