In my last post, I said how a product practice essentially allows you to make a lot more decisions to shape a product than you would normally ever be able to in just your day job alone.

Another thing I've realized while building with AI is that it is decisions, made by a human, that make a product what it is.

It's something I've been feeling lately, especially with designs coming out of Fable. At first glance, they're incredibly impressive. From a design perspective, it's a huge step up than what previous models were able to one-shot in the past. But on second glance, they all have lifeless, bland, sometimes overly-complex feel to them. They look good, but something about them just feels "cold".

To get these designs to a usable state at all, it requires a lot of handholding, tweaking, and adjustments to make them not feel that way. I currently believe it's salvageable, but only through an extensive and diligent applicationnof human judgment and taste.

This uncanny valley of AI design is something that's helped me better understand previous experiences I've had in product pre-AI around product rebuilds.

It's generally-accepted knowledge in product management that you shouldn't ever rebuild. But, in the companies I've worked at, it's just done time after time again. Setting aside all of the technical risks and problems that come from doing a full rebuild, what I didn't appreciate enough until now is how, after a rebuild happens, how completely unformed the gproduct feels.

In a way, a product right after a rebuild is a lot like an AI-output design—it looks pretty good, but it's missing all of the decisions over time that make a product actually feel robust, feel like it has a life. Before, I had felt that this kind of "dialing in" was only about the accumulation of bug fixes and business logic adjustments over. Now, I realize it's more than that—it's an intangible feel from consistent human touch that gives it character.