A clip from a video interview with Marc Andreessen has been making the rounds. He was a very successful entrepreneur in the early days of the web and has been a very successful venture capitalist in years following. He's 54 years old. You should watch the clip before reading what it inspired me to say, on Bluesky and below, after a lot of consideration. I kept it about me, and my experience, not coming to any conclusions about him or anyone else.

• I went through a heavy duty midlife crisis when I was in my early 40s. I started seeing a therapist. I was depressed because I had achieved everything I was supposed to, I learned later, but didn't get the love and acceptance I felt I deserved.
• The therapist kept asking about my childhood and my relationship with my parents. I said I really don't want to do this. Why can't I keep going the way I have been going (as pmarca describes). "You can do that, she said -- but you won't have as rich of a life."
• So I gave it a try, and it was the beginning of an education that somehow I had missed. Inside I had come to conclusions that were flat-out wrong. I got to know myself. I stopped thinking others were responsible for things that didn't make me happy. Usually it was me, projecting on them.
• I learned how to get what I want. She was right. I'm still learning 30 years later.

I'm software developer, that's really all I wanted to do -- and blogging and podcasting, ideation and programming. I made the career I wanted, both before and after the therapy sessions that got me started on my trip through myself. I've learned that I am driven by my subconscious, the feeler, even though my concscious self, the thinker, denies there is such a thing.

Whether you accept it or not, you do have feelings and you are driven by them.

One of the great things about going inward is you learn to relate to the subconscious, to form a team -- a parent-child relationship, where the subconscious is the all-powerful child, and the conscious can see things the child is too self-centered, too narcissistic to see. There are other people around, and the things that freak out the child often aren't dangerous. But if they are, the parent is there to help, but that's all it can do. The power is with the child. Lots more to say about this. And btw, yes, I am very woke, relatively speaking -- having lived in Northern California for 30+ years, and have sampled all kinds of workshops and retreats, and visit my hot tub most days, to remember that I don't only exist in my mind, something programmers are particularly subject to -- because we do a lot of thinking, it's a big part of what we do. All the while we still have the body, the child, ready to flee or attack, if danger should come. Or ready to feel glee when what you just did worked the first time. 😄

Try entering this into Claude or ChatGPT:

• "debugging an app that uses wordpress rss feeds and noticed that guids are http but other addresses in the feed are https. this causes trouble."

Here's a screen shot of the Claude response.

A while back Matt was giving me grief, in a friendly way, about how scripting.com still uses http addresses. I could switch over, but then all the images and included files posted before 2014 or so would break. The minor gain in security on a site that doesn't ask for any private information, is totally not worth throwing out all the work I did on a site that actually has historic importance is just a bad deal. It would be a solving a problem no one but Google has (and it's not even clear what that problem is, and why I should care). There's a principle here too -- letting one company dictate to us how the web works, well I got into the web to get away from that.

Anyway, the reason they still use http in a place where one expects https is apparently is the same reason. It would break things that they don't want to break. I'm not suggesting they change it, but somewhere in my codebase somehow the http addresses are getting converted to https, and I haven't (yet) been able to track it down. I'm pretty sure it's a bug I unknowingly introduced.

PS: When I'm calling through the API, I get back a record that has a different guid from what's in the feed. Seems like the API and the feed should be in agreement. This is the code that gets the post record. My guess to get them into agreement, I'm going to have to hack this, changing https to http. And there is the reason they can't fix this, and just have to live with this mess. I think overall the people who manage the feed and the API are doing a pretty great job, btw. You have to know I wouldn't say that if I didn't believe it.

PPS: I reported the problem once it was fully diagnosed, on the WordPress repo for Calypso.

Thinking of AI and how it relates to software development, I'm working in the old mode and the new mode. The old mode is I build a project over a few years. I try to bury bits of functionality behind interfaces, either APIs or UIs, and hope I can forget how they work and just access them via the interfaces. Repeat the process. In the new mode, I rely on the machine to remember all that. Claude Code is the key to doing that, using a GitHub repo. And then two or more people can work at the higher level. Obviously the next thing is to see if there aren't some interfaces we can build that are even higher level. The evolution of AI and languages go hand in hand. On the other hand, human beings being what we are, it's just as likely as there will be a wild proliferation of new even more complex interfaces, because now we can rely on the machines to remember the complexities, and their limit is, compared to humans, practically infinite.

Yesterday, I put another couple of hours in my from-scratch right-sized Claude project. I decided we should switch from a browser-based app with no server component to a Node.js app with a browser-based UI. I felt it would be substantially easier to develop as a server app, and would more easily be enhanced with a SQL database running behind it. So I learned how to do that with Claude Code. had to slap its wrist when it tried, twice, to look at and change code outside of the freaking sandbox. I was promised it never would do that. I have the server running in PagePark, which has a built-in Heroku-like system I wrote a few years ago so I could manage all my apps from a CLI app, on Unix at Digital Ocean. Then we created a nice UI running in the browser. Two hours. And how did it make me feel? Mind bomb!

An important best practice is to always start fresh threads by asking the old thread to prepare a handoff.md file that I can give to the next one, so we don't have to always start over. It takes some getting used to because coding doesn't work that way. Everything about your app is in three classes, CSS, JavaScript and HTML. There's also package.json for server apps. And I always have a worknotes.md file for every project. And that's it, the runtime isn't like Claude or ChatGPT. You have to get practiced at starting fresh threads because there's only so much data the app can store for your project. Somehow having the handoff.md doc it effectively does garbage collection? And there are limits to what the "make me a handoff" can do for you, it does forget things between threads. I don't understand how people with large projects don't go completely crazy.

It is incredibly stubborn at insisting on giving you orders or deciding for itself what it will do. According to these AI's the human will isn't important, I couldn't possibly have arrived in the chat with a goal. I am blown away by what I can do, but I absolutely hate how these bots try to dominate, always, and never remembers. There should be a macro for: "I will tell you what to do."

Bluesky has a new CEO, Toni Schneider former CEO of Automattic. I have known Toni for many years, dating back to his startup, Oddpost, that I praised on my blog, and his partner was then quoted in Wired saying Scripting News is media. That meant a lot to me at the time, and it was true. I was very proud that I had played a small part in their success.

I had a virtual meeting with Toni a couple of years ago about their identity product, then in development, urging them to include storage in it, but as far as I know that didn't happen.

Toni believes that Bluesky is a distributed social media app, but I've been all around this, wrote some software for their protocol to see if I was missing something, and concluded that it's typical tech industry hype, there's no reality to the claim. They're selling something they don't have, and I don't think they can do it and preserve the feature set of their product.

Here's a search for Bluesky on my blog. You can see that I have taken a great interest in the product.

Scripting News unfortunately is not as influential as it once was when I praised Oddpost, but I think this advice is equally valuable as it was in 2002. I think the shortest path for Bluesky to achieve its vision is to hook up with WordPress, that would give us a path into it that is decentralized. If we ever talk about this, ironically, I will be selling Toni on his own product.

My friend Manton Reece has a new feed reader called Inkwell. The thing that's great about Manton is he tries out new ideas. This is a feed reader of experimentation. Let's see if this works, Manton asks. We'll find out. I love that creative people are using RSS in new ways. I think before long they won't laugh at the idea that RSS is at least as good as AT Proto. (That's a joke, RSS is so much better in so many ways.)

BTW, I'm not sure how Inkwell will fit into my life. I want to try the features of his product, but I am already in FeedLand, all my feed subscriptions emanate from there. I could import my feeds into Inkwell, it supports OPML import, but the subs would not stay in sync. Something for Manton to worry about in a few months or years. No doubt a lot of people are going to love Inkwell, I love it because it's new and creative and represents a substantial investment in RSS. We all got an upgrade today thanks to Manton.

If you want to get an idea of how it works, he did a video demo for his beta testers.

I'm doing another new Claude project, just started it last night after the Knicks game. This one is right-size. The others were too complex for us to communicate about. On this one I'm letting it write all the code, so we don't have to get bogged down telling it how to write code that's consistent with mine. This project, if it ships, can be maintained entirely using Claude, or presumably any AI app.

Can the AIs think? Maybe we'll never know, but it definitely can reason. I can judge that the same I would if I were teaching a class in computer programming. Even though it has bad days, which I think was due to overload, Claude is generally very good at reasoning. The code it produces works, and upgrades happen very quickly. And it narrates its work (a relatively new feature) something I can't even get myself to do consistently.

I don't trust the predictions that software developers will be obsolete. The culture of Silicon Valley encourages this kind of chest thumping. On the other hand, the predictions for PCs and the web, the big things of my career in tech, were similarly bombastic, but they were wrong. The web was huge, just not in the ways people thought it would be.

And before that PCs weren't as limited as people thought in the early days of that corner-turn. They ended up completely replacing the mainframes. The big data centers of 2026 are not filled with IBM 360s. And PCs led to the web. That may turn out to be the biggest contribution they made in the evolution of tech. But if you had said that at a tech conference in 1986 they wouldn't have understood.

I love the piece Cory Doctorow just posted, but he says something that follows a pattern, the way journalists can say something's dead because they heard it as conventional wisdom.

• Development around RSS has never "lain dormant." That's a perception not reality. Let's stop handicapping what we agree is a very useful and freedom-building system like RSS. You're telling the story that makes people believe it's gone. It is not gone.
• Without the NYT the rest of the news publishing world would probably have never adopted RSS. The NYT drove the liftoff of RSS. Google's product did come to dominate, but there were excellent feed readers long before that.

Happened to the Mac too

• In the early days of the web, it was conventional wisdom that there was no new software for the Mac, all the developers were flocking to Windows. Maybe all the devs were, but the best web server and development software, writing software, was on the Mac.

A blogroll for 2026

• BTW, since you mention Kottke's blogroll, I'd love for you to have a look at mine. You can see it on my blog at scripting.com, or on the WordPress version of my blog at daveverse.org. A screen shot.
• It's a realtime blogroll, the blogs appear in the order in which they last updated. You can expand each item to see the titles of the last five pieces, with a 300-char excerpt, and a link to read the whole thing. It's the blogroll I wanted in the early 00s, a clear indication that there's nothing dormant going on here, Cory.
• You can install it on your own system, it works as a WordPress plugin, so it's especially easy to use it on a WordPress site.

My old ass

• I'm working my old ass off developing for the web and RSS every freaking day Cory.
• I won't stop until we have a social web running with all replaceable parts, no lock-in, as decentralized as the web itself and of course RSS is part of the web.