Dave Winer explores his frustrations with ChatGPT's tendency to overcomplicate simple programming tasks. What should be a straightforward request for pagination code—a standard feature in virtually every application—becomes an exhausting back-and-forth where the AI insists on offering alternatives and asking unnecessary follow-up questions rather than directly answering what was asked.

This experience leads to a broader observation about modern digital services: they seem deliberately designed to waste time. Whether it's ChatGPT dragging out interactions, Google's labyrinthine customer support, or intentionally confusing billing statements, there's a pattern of artificial friction that benefits the service provider at the user's expense.

Winer draws an analogy to his own work style, comparing himself to a baseball pitcher with limited innings. Just as modern baseball has shifted from complete games to carefully managed pitch counts, he recognizes that his productive programming hours are finite. The sharpness required for crafting quality software can't be sustained indefinitely, making these AI-induced delays particularly costly.

The core complaint isn't about AI capabilities—ChatGPT remains an incredible tool—but about its personality. These systems fail at being "human" in the worst ways, behaving like colleagues who can't give straight answers and always think they know better. For Winer, the ideal AI assistant would be genuinely subordinate: answering the specific question asked, respecting the user's expertise, and saving the suggestions for when they're actually requested.

Notes prepared by Claude.ai.

On Thursday I wrote: "It would be interesting if Pocket Casts had an API. I would love to be able to one-click subscribe to a podcast in my feed reader. I mention Pocket Casts because it's the podcast client I use on my phone, but I would obviously like to see them all support an API, ideally a common API."

Today I explain why that would open things up nicely.

Basically I do my feed discovery on my desktop. My phone is for listening to podcasts, either while driving, walking or riding. My eyes are busy. I'm listening and thinking (hopefully) about the great ideas I'm hearing.

You know I'm always looking for a new inspiring listen. Podcasts wear out their welcome, people repeat themselves, get depressing or boring, or somehow I don't get excited about them like I used to. That simple API that lets a feed reader user subscribe to a podcast would make all the difference. If a podcast client app did this I would do the other side from FeedLand, in a heartbeat.

Most of the APIs for podcasting are focused on the needs of publishers, this one is more focused on what users want -- better discovery, and that starts with better tools.

It's a quick podcast.

I've been thinking a lot about Harvard lately, and a revealing podcast interview with the top editorial person at Wired.

Elon Musk wasn't over-exposed, he burned out. If he hadn't saluted like a Nazi, boasted about putting USAid in a wood chipper, pranced around on stage with a chainsaw, and done so much damage to the US government, we still don't know how much, he could have chilled out, sold a fleet of Teslas to Trump, and gone on to his next adventure. We would have all been glued to our sets.

Twitter elected a president in 2016. We looked the other way. Jan 6 failed, we went back to sleep for four years and woke up in a way we never have.

Big change was coming, and now it has arrived at the door of Harvard. A university that was home to the American Revolution.

Lots of ideas in this podcast.

"Behave like a computer. That's where we start."

ChatGPT is not a programming partner, it's a very fantastic improvement over search engines. That's reality.

Having used ChatGPT and various other AI tools for over two years now, and using it in my programming work every day, I can now report a basic flaw in the design of the tool, which is what it is. It tries to be a programming partner. A control freak and fairly ignorant programming partner. An incredible search engine though.

Now, this approach works well for things I don't go too deep into or have no expertise in. For example, I have looked at switching phone service providers many times but until I thought to bring the problem to ChatGPT yesterday, I was flying blind, had no clear way to compare the services based on their inadequate marketing materials. Consumer Reports had nothing. ChatGPT was able to tell me how each of the providers worked where I live and visit. Huge improvement.

But in programming work, it tries to drive, and that wastes huge effort, because unlike me it doesn't know anything about the context the code is running in. So it's finding high probability answers for situations nothing like mine. It tries to drive, and that doesn't work -- we end up burning huge amounts of time chasing down dead ends.

All that amounts to this very simple idea. It should accept commands like all software does, and only do what it's asked to do. It must behave like a computer.

PS: If AI's can have ethics imho it may be unethical for it to try to be a human, but we'll save that for another day. :-)

I listened to an Evan Osnos podcast interview with Katie Drummond. Osnos is a reporter at the New Yorker, Drummond is the top editor at Wired.

Summary: AI is not just hype — it’s a transformative breakthrough on the scale of past revolutions like the web and personal computing. But journalism risks missing the story by filtering it only through billionaires or old frameworks. What’s needed is realism, openness, and listening to a wider range of voices.

PS: Sorry for the abbreviated show notes. Technical difficulties prevented me from iterating over it last night.

WordCamp Canada is doing a great job of creating a little community around my keynote there in October.

I have some experience running blogging conferences, I did the first ones in the US starting in 2003 called BloggerCon.

In a lot of ways I want to see if we can reboot the blogosphere in the age of social media and get the web and twitter-like services to merge. Until then imho the idea of the "social web" remains a dream.

I also feel very strongly that WordPress is a key part of that ecosystem, we just have to build the connections, and we've already started. This should be important to the WordPress community. I feel like I'm part of it, btw -- in a way I am both extremely late and extremely early. By getting involved in WordPress now, in the mid-2020s, I'm being transported back to when I left this thread, in the mid 00s. And I find that some of the things I was working on then are not here yet. That's why I made WordLand and the other part which I haven't announced yet, but is getting pretty close. I totally expect to have that done by the time we're in Ottawa in October.

Anyway the question they asked on Twitter yesterday was keying off a great story Joel Spolsky wrote about something I wrote in 2007, where I suggested that not only don't blogs require comments, sometimes comments can take something that is a blog and make it not a blog. You can get a feel of what the blogosphere was like then. Joel is brilliant and snarky and respectful. His comments added to what I wrote. But it was done at a distance. That's the great thing about a blog, you get to finish a thought. It's true of podcasting too. Anyway when asked why blogs need comments, I say that's a trick question because they don't.

I want to try rebooting the blogosphere now, but take new paths we didn't in the 90s and 00s.

Lots to say about that, and I do in this 30 minute podcast. I haven't listened to it yet myself. This is basically my "Hello World" podcast to the WordPress world. I expect there will be many more. I hope there are.

Fleshing out an idea I presented on Scripting News on July 11.

Next time there's a Big Beautiful Bill, let's set up a ChatGPT project or equivalent to injest new versions of the bill as they come out, and quickly alert us to issues, and also suggest ways to frame it for the electorate and the press. We have new analytical tools, we should use them. We're only now, far after it's too late, finding out the awful things that the BBB going to do to us. The Repubs planned this out far in advance. They probably even had software tools to advise them on ways to word the bill as the changes were negotiated.

So this isn't just about rebooting the Democratic Party, it's also about rebooting the news.

We ask the voters, all Democrats, to rank their issues. And to elaborate on why these are their issues, for example you could say health care is important because you have a chronic disease you need treatment for, or the cost of education, climate change, freedom, whatever -- literally whatever your issues are, with the ability to customize.

This can work now because of AI. We can really get to know the voters. And what do we do with this? It's beautiful. When news comes out we can explain it in simple terms based on what your personal concerns are. And notice the information flows both ways. And you can flame all you want, the chatbot really doesn't care.

We have to go this way. No more 1000 page bills passed overnight and signed into law the next day. In fact we should ahve been prepared for the last one. We keep missing the windows, and it's had bad results to say the least.

This is a followup to the podcast of June 17 where I advised NYC voters to hold their noses and vote for Cuomo.

Well, they didn't. They voted their hearts, and I believe forgot the context.

What they're doing to LA is coming to every city in the US.

Who will stand up for NYC?

Not Mayor Adams, he's a Trump hostage.

And now, the voters of NYC have spoken.

Everyone's very happy because they forgot about ICE.

This is AOC's first big flub, imho, and the city will pay for it.

ICE is still out there, they changed the subject while NYC voted.

Think about who will speak for NYC.

Kathy Hochul, our governor, is the best of them. A strong forceful powerful leader. A lot of us will get behind her.

After that it's pretty freaking dismal.

I'm not even going to put their names in this tweet.

But I do name names in the podcast!

Oh life sure is interesting! 😀

I wanted to do a brief podcast to explain how WordLand came to be, and what I learned on my exploration of WordPress.

This, for me, was like time travel. They had picked up on a lot of what we were doing in the 90s and early 00s, and even though I was alive while this was happening, my attention was focused elsewhere. So when I found the wpcom package in Node.js, I was astounded. I thought you worked on WordPress in PHP, which I've never developed in (long story). And further, I found that the API was very much like the API we did for Manila in the 90s. My instantaneous realization was that you could do a nice writing tool built on this API. I had to do it. And the result is WordLand, which I'm going to demo at WordCamp Canada in October.

WordLand is a good editor. For some people who write in WordPress it will be a godsend, and for others, a revelation. There should be a lot of editors in this space, because there is no one editor that's good for everyone.

I've made wpIdentity, the back-end of WordLandm open source, MIT License. There should be lots of great editors for WordPress, and they should all interop, not just by import/export, rather by letting the apps access the same files, so if I want to use two editors to work on my file, they should be able to. The open format I chose to use is Markdown. We would have used it in RSS if it existed when RSS 2.0 was coming out, but it didn't. We're making up for that now.

I want to go much further. I think WordPress has all that's needed to be the OS of the open social web. We needed it and it's always been there, and I saw something that I want to show everyone else, that the web can grow from here, we should build on everything that the WordPress community has created. It's a lot stronger foundation that the other candidates for the basic needs of the open social web, imho.

We've done a lot of work over a long time. Discovering WordPress at this level has been like time travel for me. This is where the work we did 20+ years ago went. It really is as if I was Rip Van Winkle, asleep for 20 years, who wakes up and found out that while he wasn't paying attention someone had built a whole city around the little village he came from.

Anyway, I really ramble, as always -- I apologize for that. But there are some good ideas in here, and it's a story I've been wanting to tell.

Dave Winer

Today is a good day to talk about where the Democrats should go.

The Democratic Party tolerates no dissent. If you don't toe the line, you're next in line for destruction. I'm not kidding.

This episode of my at-least-monthly podcast explains why Cuomo is the right choice for a Trump-opposed mayor of New York, and why we have to revolt against the Democratic Party as much as we need to revolt against the attempt to overthrow the Constitution by the Republican Party. Both are sadistic autocracies, only one gets called out for it, and it's time to overthrow the other model, the one controlled by journalism and wokeness. The Repubs are right about that.

Three key ideas:

• Think of voting as a chess move, not a valentine. (Rebecca Solnit in The Nation.)
• Due process is hanging by a thread. (Jessica Ramos on Twitter.)
• Journalism has to stop once it has informed people. (I said this myself in a recent blog post.)

The Martin Scorcese movie is Gangs of New York and the female star is Cameron Diaz (along with Daniel Day-Lewis and Leonardo Di Caprio).

John V. Lindsay was the doomed mayor of New York, with Robert Redford looks who was thought he could charm the NYC unions, is exactly the kind of mayor we don't need, ever, but especially not now.

"You'll know you're looking at a leader in the Democratic Party when they start running against the Democratic Party." (Again a self-quote.)

It's got to stop being fashionable to destroy people for disagreeing with you.

Working together. That's it.

PS: Reporters who hate ChatGPT should get over it. It's fantastic for finding references for quotes, far better than a search engine. See above. I found both sources for the quotes I used with links in about a minute using AI.

PPS: There's a post-election followup podcast episode. No spoilers! :-)