“Somebody can make abstract art and a casual observer can look at ... some painting of some really cool looking circles and be like, well, I could have done that. That sold for $20,000. I could have done that. It's like, yeah, but you didn't do that.”

And it's a very, it's like a much more polished, finished product than any of the Pi KVM stuff out there. Of course, you know what you'd expect. So anyway, I love the Comet Q so much. I plugged it into my data center hardware and I'm like, oh my God, I want one of these on every one of my Mac Mini locations. Right now I keep Mac Minis in three different locations. I'm like, I want one of these at least one in each location. Plus, these would have made it so much easier to set them up. So I want one that I can bounce around to. So I go, and of course, they're backordered because when they first launched earlier, I think August, they launched, they were backordered forever. I'm like, oh, God. So I'm like, I really want another one. Let me see what eBay is doing. I go on eBay and they're selling, you know, their retail price is like $140 or so. They're on eBay for like $170, $175. And I'm like, oh, I do, I really could use one more at least. So I can move it around my server. It's like, let me get one more on eBay. And I'm bidding and I bid all the way up. And at the last minute, somebody sniped my bid for like $175. I'm like, oh, geez. The next day, they're in stock at Amazon for regular price. So I ordered five. I'm just like, just go. I just want so many of these. And as it turns out, I don't quite need five. That's why I gave one to Casey. But I will mention they are at this moment still in stock on Amazon, actually on sale for only $119. And if you are a nerd who needs a KVM of any kind, like a remote KVM, because the other thing too is like all the Comet series of KVMs, they are themselves able to run tailscale clients. So you're able to connect it to your tailnet and then you can from anywhere log into this KVM and whatever the machine that it's controlling is, which is very useful because the KVM itself is a tail scale node. So once you're in like the tail scale ecosystem, if you manage any computers remotely, especially if you're doing like, you know, like a Mac mini in a closet kind of situation or, you know, or somebody else's house or a data center, something like this can be awesome to save your butt if something breaks screen sharing. Or if you don't want to leave screen sharing open for maybe Mac OS security hole reasons, you can just not even have screen sharing enabled and just use this as the KVM. You know, obviously you do your own security research and stuff, but like it's such a fun little product. So I can strongly recommend like if you're nerd enough to listen to our show, the odds that you are nerdy enough to want one of these, I think is pretty are pretty high. So I would snag one while you can. Anyway, so yeah, thank you to GLINET for sending the original freebie and I bought five more and gave one to Casey. All right. Now, you just were disclaiming your. Your AI coding work. And I know, look, I'm going to piss off everybody. First, the Tesla people. I'm sure they're already writing in. And now everyone else who doesn't like Tesla, they tend to hate AI. So now I'm going to piss them off. Sorry about that. I'll do my best. But you said that, like Lex was saying, that being the director of the AI is where the value is now. And the more I use AI coding tools, the more I see, it is obviously not the same thing as making an analogy to assembly language versus higher-level languages. It's not quite that. It's a larger and more profound jump than that. But the reality is that AI generally is not very good at figuring out what to do on its own. It is, however, very good at executing on your ideas. And it's also not very good at doing a lot of polished, nice behaviors. It's not very good at coming up with thoughtful features. It's not very clever. It's not very good at like making nice designs. And these are all things that you as the human can be responsible for making and directing and coming up with. What AI is really good at is kind of the worst parts of programming to a lot of people of like, oh, I have to build this stupid form and I have to debug this really tricky thing. Like it's really great at that. And that's the parts that we kind of suck at. It's really bad at the parts that we are good at of like, hey, what if I had an app that did this? Like the idea parts. And as like, first of all, as good idea people, but also as experienced programmers, we bring skill to using AI that the average person does not have. And that skill is still valuable. And I don't think we need to be sorry about using AI. I think using AI is helping us deliver great software, but we are the ones still coming up with the ideas, figuring out what's even good to do, and then figuring out all the things to ask it to do, everything to ask for. And then when it gives you the results, we are the ones that are like, this button should not be here. And this is, hey, this feature kind of sucks. Why don't you move it into this kind of navigation? Or this should really be redesigned to fit this other pattern here. We are the ones who know what to ask it to do. And if you look at the AI-assisted coding output as a product, what you or I are able to have AI make with us is so much better than what a random person who has no experience developing software is able to make because we know what to ask it to do and we know what kind of standards to hold it to. People who are like, oh, it's useless, like anybody could make that. It's almost like when, like in the abstract art world, somebody can make abstract art and a casual observer can look at some, you know, some painting of some really cool looking circles and be like, well, I could have done that. That sold for $20,000. I could have done that. It's like, yeah, but you didn't do that. And the person who did that, their artistic contribution was not that they were the first people to ever think of putting circles on paper. They put those circles on paper in that way. They had that idea to do it at that time with those details and those methods and with that kind of story around it and that kind of marketing. And all of that, those are all really important parts of app development and product making. There's a lot of people out there who just keep one-starring everything I make. Oh, he said he made it with AI, so it's useless. How dare he charge money? And these people, look, I'm probably going to be dealing with people for the rest of my career. We all will. But the reality is what people want is good products that are thoughtfully made, that work well, and that solve their problems or delight them in some way. It doesn't matter who's writing the code, whether it's some AI model churning out the actual code behind that doesn't matter at all to the customers. All that matters is: do we make something good and does it work for them or to make them happy or both? And if AI helps us do that better, that's great for everybody. I honestly, I am past the point of apologizing for using AI. I don't think any of us should really be apologizing anymore if it helps us achieve those goals. And I know that's going to be a little unpopular, but the goals speak for themselves because what almost all those people also want is for us to fix bugs in our software or to not chip with them in the first place or to work on their pet features. They want us to have the skill and bandwidth that AI gives us. But they just, you know, they're uncomfortable with some of the reality. Of that. And I think, look, there's only so much we can do to keep those people on board. I'm sure I've already lost them. So sorry. And I don't mean to be flippant about this. I mean this really seriously that like the skill of making software is now involving using AI agents to help you. And that results in more, better software. So that's great. And I don't think we need to be apologizing for that.