The moats, then, exist where what constitutes tight tuning is non-trivial and often non-obvious. One general piece of this is that it is more difficult to tune something over long time scales than short time scales. So software that has “nothing but immediacy,” like Photoshop—apply the filter, there it is; crop the image, there it is; lasso that thing, there you go—that has no moat. It’s too easy to reconstruct. On the other hand, things that contain progress, memory, projects, compounding impacts and effects—these things are non-trivial to envision and to test and to optimize. These things require long-term coherence. You see it in the demos all over Twitter. A little applet here that’s kind of neat. A beautiful movie there that’s brief but compelling. Linking these things together into a longer arc that has an internal coherence and connectivity to it—this is decidedly an unsolved problem. And in many ways, it’s not a problem to be solved at all. It’s where the art lives. It’s where even things like wisdom show up: what matters and why. We’ve all seen this idea that what matters most now is taste, but that doesn’t quite hit it. Wisdom is closer. You have infinite potential. Okay, then here come the constraints of time, attention and value. What matters and why? What shouldn’t happen? What constraints should be set? How do these things internally relate to one another? Yes, as AI continues to advance, it will ape some of these things more and more. But there is an ever-receding horizon where only something you might call wisdom can live. Taste is too easy to feign. Taste is the wrong word.

that* doh