Was working on the boy's voice from the "Go Forth, My Unstoppable Machine!!!" log and, as you can now see in the log itself, Claude was generating HTML artifacts to show me the various voices and possible variations. Eventually I said I was going to the gym and that I would continue iterating with him from there. Claude then took it upon himself to create a version of the HTML artifact that was usable from a phone, with bigger buttons and better layout and so on. I was so impressed by that. At this point, this is extremely simple behavior, but it still never ceases to amaze me that computers can now do this. I can say "I'm going to the gym," and the robot automatically reasons that I'll be using my phone and that it should build the artifact so that it's actually usable from a phone. It's just so... it feels like magic, like everything else about this technology. Like this tweet:

It's all just very impressive. Not to mention the actual content of that particular log, which is the robot training an entire new voice model based on a selection of voices to get the exact boy voice I want. I have no idea how long it would have taken me to learn to do that, but I know for a fact I just wouldn't have learned it because it seemed an insurmountable task, and it does it just like that, this is a case of ∞x productivity gain. I can't help but feel amazed at all this, I thought at this point I'd get used to it but no, it still keeps surprising me.