I would say AI has been around more like 8 months. The 2026 models have been a serious game changer. AI from 4 years ago was interesting but not very effective. Last year at this time it was useful but not a game changer. Only 2026 has started giving us models that are causing upheaval how we work. And even with AI, it takes time to swerve legacy code bases with millions of lines of code and still require thousands of hours of real-world testing before they can be released to the public. But it's coming. Oh my yes. Bugs we've had that have been a problem in the codebase for years, because no one could quite nail them down, are being solved now in a sitting by Fable. Now think about how this is going to hit robotics. We've had robotics for decades. Clumsy, custom, hard to build, very specific jobs, but applying modern AI to robotics is going to give us robust systems that can learn to do quite literally anything, and will no longer takes us 10 years to build. I think we're going to see some wild shit over the next 5 years or so. Production will still take time to ramp up, and who knows what regulation will do, but I think 10 years from now, you will call for a plumber, and you're going to get a couple of robots that come out.

He paid $700. Some goes straight to the state in the form of sales tax. Some goes to the state via the restaurant's income tax. Some goes to the state via the income tax on the employees who get salary or hourly wages. "No tax on tips" but the server who got the tip will eventually spend it, sending that back into the economy and more rounds of taxation. We already have a very effective tax on wealth. What we should really be doing is encouraging them to spend it.

Ask claude to audit your skills. Try the built in /skill-creator. I have found that Claude will add to skills in such a way that it is breaking the skills as it writes them. In particular, it tends to put too much into the description, the description becomes too long, and now it won't read the whole thing. I think you can also look into breaking skills down into multiple playbooks. Rather than one giant skill, it's more like a skill header that then directs it into individual playbooks, each of which is focused. (Assuming it's a complicated skill, anyway.) I've also had good success in doing this: You want to use skill X. Open a Claude session and tell it to open a subagent, and that subagent is going to use skill X to do a thing, and the top level agent's job is to audit the subagent and observe how it is using the skill, then make recommendations for how to improve the skill. This is another aspect of AI that I think will get cleaned up over time but for now, conducting skill audits to make sure they are working the way you thought they were working is something you should do now and then.

The funny thing is that artists themselves would be the best users of artistic AI. A non-artist can make it happen but someone with real training and talent could do even better. It's the same way with programming. A non-programmer can definitely make a web-app now that does what they want, but a programmer could use the same tools and make one that's better, more full featured, and works better where it counts.

Alterac Valley was what the entire game of WOW should have been like. I'm still mad that they took a great wargame and made an MMO like "Oh it's later the war is over, grandma lost her necklace, go fight alligators to find it."

Probably need to think about spinning up your own management tools then. Have Claude make a kanban that helps you keep things organized. What is being worked on? What is in flight right now? What's done? 🤔 I've seen examples of others doing this and talking about it here makes me want to try it for myself. I do find it hard to track more than about 3 things at a time but that's because I'm flipping between 3-10 different tabs (3 things in flight, 0-7 more that are paused, or possibly done, but I haven't closed yet just in case they need to come back). It can take me a minute to figure out what was going on when I flip back to something. Creating a better interface would ease the cognitive load.

What keeps people poor is either the unwillingness or the inability for self-improvement. What we really need is for it to be cheap to switch careers with a change in education but, ironically, education is immune to capitalist pressures, which is why it's extremely expensive.

And yet, I bet that given the choice between buying American, and supporting American workers with good salaries and good benefits, or buying made-in-China, and supporting literal slavery, but your t-shirts are $10 cheaper, you support slavery every time.