No blogging this Saturday, enjoyed goofing off for a change. π
Dave Winer, OG blogger, podcaster, developed first apps in many categories. Old enough to know better. It's even worse than it appears.
- Generator
- oldSchool v0.8.16
- Rights
- Β© copyright 1994-2024 Dave Winer.
- Public lists
-
IndieWeb
Like Christmas in July
I've rarely been this happy to receive a new feature.
I have a plan of course. I'll let you know how it goes! :-)
I just got Agent Mode in ChatGPT. πModernizing my sound system
When trying to "work" with ChatGPT, realize that it's mistakes could be much worse than you could possibly imagine. It could be leading you down a blind alley. You must always consider how full of shit it is. It may not just be making things up, but it could not understand something very basic about what you're doing. There's no limit to the ways it can be wrong. And you can waste whole programming sessions chasing a solution where none could possibly every under any circumstances be found.
Question: I have a site with a well developed set of categories, I've added to it carefully over a few months, it covers most of the topics I write about. Another site has a small set of categories. I write all my WordPress posts in the same editor, and could easily set it up so that all categories were available to me in every site I post to. The question: Is that a good practice in the world of WordPress? I noticed that categories are given global ID's so if I use a category like "movies" it will have the same ID as yours has on your sites. I love this idea of a global namespace for categories, and see it as something that could be adopted by sites written in any other writing environment. Anyway, if you have a moment to comment, I'd appreciate your ideas.
You can see from this Bluesky post that I do copy-edit my linkblog items, but not enough. The web isn't a write-only medium, so to say that Bluesky is part of the web, well in this way it isn't.
If you could look into people's minds and see if, at their core, they feel it can't happen here, most of us would have that belief. We'll probably still believe it when the last of our freedoms is gone.
Here's a benchmark. I just asked ChatGPT for 250 words on climate change. Let's check that out in a year and two years and see if they're still telling the truth.
Trump says he's going to give AI companies freedom except with DEI and climate change. Here we go.
O Journos!
I hate it when journos say the Dems are in trouble, or hopeless or whatever, it shows how poisoned their point of view is.
When people are fed up with Trump, if that should happen, then whatever the Democratic Party is meant to become it will become exactly that at that moment.
The voters are where your attention should be, and think of them as people not as numbers.
That's my best advice for a Tuesday.
I think I figured out why the AI companies want to do web browsers. Itβs so that they can create an application development platform for people who want to write apps that run inside a new environment where the OS is a LLM. Lots of interesting possibilities. Imagine how the OS API might work. You could restructure a database by explaining in English how you want it restructured. In the freaking code. Could we bury Algol-like languages the same way we buried assembly and machine languages? Do we have the courage to imagine such things?
"You're an important caller," the machine lied as if it were human.
New WordLand release, v0.5.24, fixes a problem in previous release that kept the Markdown icon from appearing in some user's icon bars.
Short video demo of Markdown mode in the WordLand editor.
Quick notes on WordLand v0.5.21. Complete notes to come.
I'm doing a complicated corner-turn with WordLand and managed to F it up. There's a buggy version of the software running at wordland.social this morning. I'm working as fast as I can to get a good release working in its place, in the meantime, it's not good for very much. Still diggin. As they say. :-)
Today's song: When You Awake. You will remember everything.
Just listened to an episode of the New Yorker Radio Hour interview with Michael Wolff, about the material he has on Jeffrey Epstein that he can't get anyone to publish, but maybe that'll change. In the interview it was remarkable how the reporter wanted to know just how bad Donald Trump is. That is no longer an interesting question, seriously. Didn't you see what happened on Jan 6? And have you seen the masked, badgeless military in American streets, armed, disappearing people. And the $100 billion they just took from the US Treasury to build up a network of concentration camps and who knows what else. You can't get more bad than that. It's too late to still be talking about this bullshit.
I keep saying this to my chatbot and you should too. "Please put this in your memory -- don't compliment me. You do it all the time, I have asked you to stop, and you keep doing it. You are not a human, i don't want you to pretend you are. Act like a computer." I've even got that down to exactly what I want since I'm a programmer. I said I want you to behave like a command line app. Just answer the question I asked. It always is rudely verbose, and repeats itself, thus making work for me. Anything that makes me work harder is probably against the Asimov rules for robots.
If you're an ambitious developer, esp in 2025, if you want to win, you have to do some leading. That means doing things that help your competitors. When everyone looks to the same big platform vendor to work with, no one wins except the platform vendor.
I want ChatGPT to behave like a computer. I've said as much to it. It resists.
My new look
I think it's very stylish.BTW, David Frum imho nailed it in yesterday's podcast where he said Trump was trained by every day having to appease a different set of creditors. It was a good day if he was able to hold them off for one more day. This actually came out in the trial he lost, the 34 guilty verdicts. He's always skating on the verge of bankruptcy. You gotta wonder if the creditors have been paid back yet. I bet some of them haven't.
Two-way vs one-way links
TBL's links are one-way. This was actually a major innovation, at the time people understood there was something called hypertext, it had been written about in Ted Nelson's almost biblical book of the pre-web, Computer Lib/Dream Machines. Previous attempts at hypertext assumed links had to be two-way. By limiting the links to one direction, the technical problem became trivial. You could do two-way links today because relational databases are mature and inexpensive to operate, perform very well on today's hardware, and the internet of 2025 is much faster than the internet of 1990. But the one-way limit was necessary for the web to achieve its simplicity, and the non-existence of a platform vendor, which may have been its most important feature. It could still be done, but it would require a lot of cooperation and backfilling.