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.
# Mon, 09 Mar 2026 21:15:30 GMT ## Bluesky: "The reason we have enough money for a war is that we get to print money because we have the reserve currency that the whole world uses. So we could afford to buy you a house or pay for your healthcare or forgive your student loan debt but we don’t because I don’t know." # Mon, 09 Mar 2026 14:33:46 GMT ## An app I'd like someone to do. I want to underline the word reason in a blog post I wrote, below. I want to point to a page with a definition of the word, as a verb, not a noun. As far as I can see there is no page on the web for that. Your app will have a dialog at the top of the page where you type the query, and it generates a page with a static URL that I can point to where the definition will display if the user clicks my link. I would paste the URL where I want it. And that's just the start, the key thing is short replies to queries needed to support something you're writing. I'm surprised Google doesn't do this. And I'd much rather use someone other than Google, but it has to be someone who will be around for a while. You can put an ad on each of the pages, but don't overdo it, or you'll incentivize a competitor. # Mon, 09 Mar 2026 22:23:20 GMT ## Manton's Inkwell -
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.
# Mon, 09 Mar 2026 14:08:58 GMT ## AIs can reason -
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.
# Sun, 08 Mar 2026 16:30:34 GMT ##
Let me tell you something about AIs. They are not in any way ready to develop the kinds of apps I make. I spent a full week trying to get it to do so. What happened here is that we all were blown away, correctly, with what ChatGPT could do, and loved that it kept getting better. I've used it and Claude to make apps, and that is also amazing, unprecedented, maybe the biggest innovation ever. But. It doesn't have the memory you need to keep a full app in memory at once. And the tools we have now, compilers, editors, runtimes, do remember the whole thing, they are really good at that, but they don't understand at the level a human can and does. And sessions are too limited. And it makes unbelievably huge mistakes. Maybe they will get there, but we also had high hopes for the last breakthrough, the web, in its early days, and it didn't achieve its promise. Turns out the web gets you Trump, and Trump just discovered he has nukes. Cory talks about enshittification and that's right -- but it's even worse than that. The tech industry always oversells the innovation. I am one of them in that regard. In this one I'm so far just a user. Also I haven't given up. Still diggin!
# Sun, 08 Mar 2026 16:11:09 GMT
## Carville is obviously right. No political party can afford to demonize a group of voters based on gender and race, esp when they make up approx 33% of the electorate.
# Sun, 08 Mar 2026 16:15:32 GMT
## 
Reporter at the Guardian: "We don’t talk enough about how morally depraved the tech industry turned out to be. Every single ounce of their self-regarding statements of values was an outright lie." It's true. I was covering tech realistically starting in 1994, was writing for Wired, people thought I was being too hard on them, but I was actually like you too easy. But people didn’t want to believe tech was evil, they believed that the young people that were running tech were idealists and maybe they were when they started, but by the time the billions started flowing and they stopped caring about people and started only caring about money. A piece I wrote in 1996, after going to a tech industry conference."
# Sun, 08 Mar 2026 16:29:43 GMT
## He was a trust buster
- 
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.
Happened to the Mac too
A blogroll for 2026
My old ass
I remember liking the first three seasons of Industry on HBO, so I just watched them again. It's a Succession clone, in a way, not exactly the same story, but the same type of story. I waited until the final episode of Season 4 had aired to start at the beginning. So now I'll be watching fresh stuff, which is kind of scary because I found that I had forgotten some of the big plot points, I wonder how much of the new season I'll understand. I also found it dragged toward the end of Season 3, where they do a trick with the audio, make it sound really portentious and dramatic with a promise of evil, for events, which without the music would seem mundane, tiresome, kind of pathetic actually, embarrassing and just plain stupid. But at least it was just part of one season, there are some series that are all about nothing, made to seem important. I try to imagine the writers' room at such shows. Do they know how ridiculous it is? Maybe they don't care. Next up is The Pitt, which everyone says is great, esp doctors, tried watching it but couldn't stand the gore.
# Fri, 06 Mar 2026 19:59:48 GMT
## If you have an X account, esp if you have a lot of followers, please RT this post. I'd like to get my real account back. Thanks for your help.
# Thu, 05 Mar 2026 23:05:40 GMT
## On the other hand, it's hard to get Claude.ai to really apply itself to my own software. It likes to drive. Same with ChatGPT.
# Wed, 04 Mar 2026 14:16:25 GMT
## 
The thing that's amazing about Claude.ai is that it understands how software works. I can talk to it about software the way a football coach would talk to a player about football. I gave it some instructions in English about how the outliner was going to evolve. I asked if it remembered how Rules worked in MORE. Yes, it explained it correctly. Then I said I'd like "faceless" rules, where we could edit the source so the outlines looked the way we wanted them to look, using Rules. In the time it took me to write a sentence here, it finished the job. I added a home page for the AI outliner folder with links to the other docs in the folder. Then I did a bunch more changes, I could go on like this forever. It was like working with a team on a product, only the team turns around new versions in seconds, and eventually runs out of space (gets tired?) and I have to start another thread. I just did a transition and it seemed to pick up pretty close to where we left off. I have a lot of ideas here. Expect an explosion of new versions of popular software writing by individual people. We'd better make sure the standards of the web are really well documented.
# Thu, 05 Mar 2026 03:58:10 GMT
## What if friends treated their friends as nicely as they treat dogs. When you sensed they needed a little support, you'd look them in the eye and say "Who's the good girl?" Rub behind the ears. When they sit give them a treat. Inside of us, everyone, including you, is a little pup who just wants to know they're in the right place doing the right thing.
# Wed, 04 Mar 2026 21:05:59 GMT
## We had a problem today with one of the servers, it meant a bunch of services weren't working. Never found the actual problem, but something changed and the misbehaving server started working. Learned a lot about managed databases on Digital Ocean.
# Tue, 03 Mar 2026 14:54:55 GMT
## 
I asked Claude.ai to "write me a nice little spreadsheet program that runs in the browser." Here it is. It looks like a spreadsheet app but it's missing most of the really good commands, like defining the value in one cell with the sum of two other cells using point and click. If you go down this path, ask it to keep a user's guide current, and then ask it to put in features, and just describe them in standard spreadsheet terminology. The trouble starts when you want to make something that doesn't have a standard terminology yet because it's new.
# Tue, 03 Mar 2026 18:33:27 GMT
## Then I had to ask Claude.ai to write me a nice little outliner that runs in the browser. And it did. With a flourish. It was designed to make me the guy who designed outliners for most of a lifetime, and I have to say it was very nicely done, for a two-minute project. Even for a two-week project it's pretty nice. Then I asked it to do a priorArt outline, and it looks really good in the this.how template. The power of standards. And I had a full day of work even while Claude.ai was doing these mind bombs for me.
# Tue, 03 Mar 2026 18:57:59 GMT
## I asked for a feature of the outliner from Drummer that it automatically opens a file in read-only mode if there's a URL parameter with the address of an OPML file. Like this.
# Tue, 03 Mar 2026 19:22:57 GMT
## Really Simple pizza
- 
I like the way they organized today's Fediforum conference. (They call it an unconference. I use the term to mean something very different, and we used it first at BloggerCon.)
They asked for "position papers," and chose a set of them to be presented.
Inbetween, they had a set of virtual tables where six people could join and have a conversation.
It wasn't boring. And that's the first requirement for a conference.
Some of my takeaways from the meetup.
If you followed me on Twitter, please follow me on Bluesky or Mastodon. As far as I'm concerned Twitter is gone. Not because I'm religious about this stuff, but my account got hijacked and I can't get it back, so let's close that chapter. It was a great innovative product that also held back progress on the web for 20 years, and it made some people I knew a long time ago fabulously rich, and it would have been nice of them to not do this to us, but what the f, it is what it is. One more thing, guys -- pay your taxes.
# Sun, 01 Mar 2026 15:35:30 GMT
## A bit of general advice about using ChatGPT et al, never let it rush you. You do the thinking, it does the stuff you ask it to do. If you're not careful it'll quickly start giving you orders.
# Sun, 01 Mar 2026 14:33:12 GMT
## opmlProjectEditor format
- 
Some time in 2013 I started editing all my JavaScript projects in the Frontier outliner, and in doing so I designed a format that could contain a whole project. And it worked, I continued building it, and to this day I edit all my projects in this format. It does a lot of work for me automatically, making it possible for me to build more complex stuff.
It turns out you can put a lot of code into an outline on today's computers. The outliner in Frontier was designed to perform well on a 1990 Macintosh with 1MB of memory, so you have to do a lot of writing to overload it.
I am doing a project with Claude.ai which I'm editing of course in OPE format. So I had to teach it how they work so I could give it one of these files, and it would not only be able to understand it, it could make mods and send it back to me in the same format, and with the code more or less formatted the way I like (still working on that).
Yesterday we started the project. I asked Claude to document the format which I called opmlProjectEditor format, which I am now publishing for future reference by myself, other AI bots, and anyone else interested in this.
Here's a link to the opmlProjectEditor docs on GitHub.
I started a this.how page so I can add more links as this develops.
Every source.opml file in my projects on GitHub is in this format. Here's an example file in OPML, and here's a link that opens the file in Drummer to give you an idea what it's like to work in this format.
# Sun, 01 Mar 2026 14:11:59 GMT ## Archive for Scripting News in February 2026, in OPML, as always.
