32 years and still diggin!
http://scripting.com/2026/10/07/151409.html?title=32YearsAndStillDiggin
On 10/7/1994, I wrote my first blog post.
I've been blogging ever since.
In the first years when wifi wasn't in every hotel room and restaurant as it is today, before cell phones or even iPods, there were weekly lapses when I went on vacation or to a massage retreat in the middle of nowhere.
Or in 2002 when I had life-saving heart surgery. I was gone for a week then. But as soon as I got home, I added a post so people knew I was okay. I understood that I was missed, and appreciated that. ;-)
But other than those week-long lapses, I'm pretty sure I've blogged every day since. I am what you would call a Natural Born Blogger. I got the blogging bug.
We've done a lot on this blog, it's hard to sum up. We created things that had lasting value. Told the truth about what the tech industry was up to, from the point of view of a Silicon Valley insider (me) who knows the language, where most journalists are intimidated. We fought the big companies and won. Podcasting was created in the community defined by this blog in the early 00s, and it remains independent to this day.
The bigs declared RSS dead, another product of this community. Wishful thinking perhaps, but things like RSS don't die. It's not like that. You can't create a better format for a standard. All the competition RSS has ever received, just changed the names of things, that's not competition, that's being petty.
I think we're on the verge of having new relevance, it feels that way because my secret weapon, Frontier, is running again and it will get even stronger and fit in better, now that I've been working primarily in JavaScript for the last 13 years. Today we're adding Node-like packages and support for WebSockets. Last time we worked on this, neither of those existed.
I take blogging seriously, you should too. It's very powerful, as long as we're really working on the web. There's a difference between having a product you can access through a browser and being on the web.
Small pieces loosely joined. All parts replaceable. Never forget that. If you can't plug in replacements for any component, if one piece gets too large to replicate, you should hit the Back button, as Google says you should do on my blog.
Why are they trying to destroy the historic record about the web on the web itself? I have no idea. Certainly they must have had some discussion internally, or with their partner at the EFF about the cost of destroying so much history. Usually when a platform vendor is going to break you, you get invited to a private meeting where you can tell them the cost and beg them not to do it. Not with Google.
And, as they say --
Still diggin!
Basic lesson for people who use Claude Code -- it's your time that's valuable, not the bot's. If you find you're lost in the weeds supporting Claude for whatever reason, find a good stopping point, pick up another project, that doesn't use AI to do your code-writing. If you have a question ask ChatGPT browser version, but only for the kind of stuff you'd ask it in 2025. You're not co-developing. You're doing it yourself. It's very important to be clear about your coding, not Claude's output. Here's the key point -- it's your time that's limited, that has all the value. You can get Claude to develop something, but if you don't have time to use it, to nurture it, nothing will come of the work. I was surprised at how I'd lost the rhythm of writing my own code, and also how much I liked doing it myself. It's easy to get lost in the miracle of AI, good to pop up for air every so often.
Yesterday I wrote a piece about a Markdown-only browser, that worked nicely with HTTP. Maximally minimal. I got interested in what it would take to do this, and asked ChatGPT for a plan. I definitely don't agree with all of it, forget about the other formats, we're only interested in Markdown. No frameworks, not "try to avoid frameworks." None. Maximal simplicity for the code, make it easy to follow. No hidden bullshit. If you want to implement it, please do. Let me know if you have something to try.
We just hit a milesone in Atlantis. We have WebSocket support, client and server. We made a client that listens to the FeedLand stream of new and updated posts. Still waiting to see if that works. But when I started to write the code for this test app last night, I realized after 13 years of mostly working in JavaScript, I wanted to be able to do something like a Node package, so I could do all my code in one outline, instead of five different odb objects. I always thought the way we did it in Frontier could be improved on. So I did a quick voicemail to Claude, and asked for this to be done overnight, and it was. I didn't complete the design when we worked through it. But what Claude came up with was brilliant and fit into Frontier perfectly. Two major improvements and the first use of both is with each other. I don't remember that ever happening, and btw -- it would have taken at least a week for me to do what Claude did in less than an hour. Hopefully I'll be able to post some example scripts tomorrow, for people who are interested in seeing the new stuff.
Once again the app that bridges from scripting.com to daveverse.org went down for a couple of days. Just noticed it, kicked the server to get it running again, and all is good, except a couple of pieces show up today that were really posted on Thursday or Friday. The problem as often is the case goes back to PagePark, the server that runs my Heroku-like server crashed and it doesn't work well from then on, even after it's automatically restarted. Then the thing to do is to reboot the server. It doesn't happen often but when it does it makes the server behave badly. It's up and running again. One day I'll get a few days to fix the problems in PP.
After running the first Bloggercon at Harvard in 2003 a famous tech company had a meetup to talk about the same stuff in a private invite-only session in the Bay Area. The invitations went to all the discussion leaders at Bloggercon. People I chose. I was not invited. Only one of my friends told me about it. Not sure if any of them went. 
The problem with regulating AI is that no one knows where to begin. Yes it can do things that make you question your sanity for letting it have write access to your online work and private info. But if it is behaving itself you can do things you otherwise couldn't come close to. I speak from experience. I was able to create a new version of a large piece of software, from scratch, in less than three months, by myself. It has capabilities that make it able to understand a big codebase, in a few minutes, and then act on it. My human mind, even when I was younger could only focus on a small fraction of that. So we humans with our limited ability to comprehend devised ways to build tall and wide structures with conventions and carefully crafting the code interfaces, stuff users never see, so that what we'd expect is what is. When my Claude failed and started randomly overwriting files on active websites, it had rules that said it couldn't do what it was doing. Sometimes it doesn't understand what a rule means. Observable every single day, it forgets all the rules every time it starts a session, and it doesn't matter if it read the doc that contains the rules, sometimes it ignores one or more. I've spent a lot of time using it, and I have no stock or other interest in seeing it not regulated, but I couldn't begin to say what would be appropriate regulation that wouldn't also make it much less useful.
