2026 Gallery
12:16 Saturday, 18 April 2026
Current Wx: Temp: 70.32°F Pressure: 1011hPa Humidity: 59% Wind: 18.66mph
Words: 154
Trash got dumped and the meeting went well. There are so many details to keep track of when you're building a house!
I got back to ChatGPT and Gemma and managed got the page up. This is all Gemma's code, but I had to create the JSON file listing all the images. It still needs some work with the arrows, but you can use the right and left arrow keys. I need to figure out how to add captions or titles, but that's for later. This is pretty cool for now.
A cold front is supposed to blow through later, and the wind is picking up. I think it's supposed to arrive about the time Mitzi's plane is supposed to arrive. That won't be fun.
Our Florida neighbor is coming by in a few to pick up a package he had delivered here. Other than that, not much more on the agenda for today.
✍️ Reply by emailCollaborating
06:43 Saturday, 18 April 2026
Current Wx: Temp: 54.18°F Pressure: 1013hPa Humidity: 85% Wind: 8.43mph
Words: 556
I've got to take the trash to the transfer station and I have a meeting with our builder at 9:00, but I think I have time to post something.
I installed LM Studio yesterday, then Gemma 4 E4B. It's not the largest model I can host, but it's a starting point to figure out how useful this may be.
Here's the project: This is something I've been wanting to do for a long time, but it just felt like it was going to be too much work trying to figure out how to do it. If developing web pages was something I was going to do for a living, or a lot even, it'd probably be worthwhile to study html, css and Javascript. But I don't, and I'm not and I'm disinclined to even get one of those web page editors like RapidWeaver. (Which I actually have a license for. I used it for my futile campaign website several years ago.)
All the images I post here are in a folder on the server. I'd like to have an annual gallery where you could click through all the images I'd posted in a given year. Pretty trivial stuff, I know. Just, not for me. But maybe Gemma 4 E4B could walk me through it.
ChatGPT thought that was a very appropriate task for that size model. So I wrote a prompt and asked ChatGPT if that would be appropriate for E4B. The AI refined it to be less narrative and more bullet-points. It also added a coupe of technical points I'd omitted because I didn't know any better.
So I copied that over to the prompt window in LM Studio and E4B began crunching away. It spit out an html page and I copied the result and pasted it into ChatGPT for its review. It declared it was a "solid result...better than average for a 4B model." (Was it throwing shade?) And then it walked me through a code review.
So I have some feedback to offer E4B, but first I want to solve an infrastructure problem. I need to have the list of file names in the folder as a JSON file to be processed. So I'll work on that with ChatGPT and E4B, and then return to the gallery page.
I also have to modify the AppleScript that creates photo posts in the marmot. I'll need to add a step for it to append the photo filename to the JSON file. I'll muddle through that with some assistance from either ChatGPT or maybe E4B. I don't know if E4B knows anything about AppleScript.
There are at least two challenges to overcome in doing anything significant with a computer. That is, the kind of thing that might make a computer a "bicycle for the mind." The first is knowing what you want. The second is figuring out how to do it. An LLM can help a great deal with the latter. It can also help a bit with the former, because you must know what you want to write a meaningful prompt.
Anyway, looking forward to it. An LLM is a much more patient and responsive interlocutor than someone in a forum, assuming your "context window" is large enough.
With that, I guess the trash goes out and the beat goes on...
✍️ Reply by email