Ah, that’s a limitation of the web component at the moment – it only shows tags from the server that’s hosting it. So if you changed the source to point at the main server, it would show tags from that server. But not both yet. That feature will be part of future updates. I clicked through your blog a little and things look like they’re showing up properly: https://octothorp.es/get/everything/posted?s=keysklubhouse.com
Oooh I see! Well that’s cool, I’m glad it’s working as intended and I look forward to future updates! 
Is it a good idea to host a webring with y’all now or wait until 1.0 comes out?
It has been stable for the several months I have been using it, which is longer than a good number of web rings I’ve seen lol
1 Like
niceee! ive just registered 
lol ok great thank you
Hi! Yes, I would call us stable! And the actual stuff you put on your site should stay usable through any future development or API changes. Lemme know if you have any questions.
1 Like
Whoah! This looks super cool 
Thank you for setting it up and we adore the terms you set out for the website!
2 Likes
Out of curiosity @nim , do you have any analytics on what tags people tend to visit / search? Rather than which tags have the most posts?
I have been using the tags for a while now, and it has occurred to me that I have no idea if anyone is clicking on them to explore further, or finding me through them haha
1 Like
Wahoo! I’ll try to set that up… Once my computer is fixed… Lol
Interestingly we’ve kinda built things in a way that hides a lot of that traffic from us. We can see when someone comes to our site and views a list of tags, and the Weird Web October tags are at the top of that list, since the WWO site links directly to them. But, if someone puts the web component on their site (the little hashtag that opens up into a list), we don’t have a way to track that interaction unless we deliberately put trackers in there, which isn’t really in the spirit of what we do. And besides, anyone can embed a list of tags on their site to get around that. We could add a utm value of like "via-octothorpes"to those links so that you could see how many links are coming in to you via an OP list. But I’d probably still want to provide that as an optional setting.
1 Like
I realized only after that it is kind of impossible to track users without tracking users 
haha but thanks for clarifying
It also doesn’t help that it looks like at some point I changed my templates and removed the links to the tags (whoops!) So I probably arnt sending people down tag rabbit holes myself anyway
Unless I’m much mistaken I’m seeing extensive AI (Claude) usage in this project, which is a bit of a deal breaker. Sad to see this stuff in what otherwise would be a rather interesting personal Web project.
Sorry, can I ask where exactly you see that? I don’t see anything obvious myself and it feels like a weird accusation to throw around
you can see that claude has ‘co-authored’ some commits in the repo, ex build: make the production Vercel deploy build on Node 24 · stucco-software/octothorp.es@9f57b67 · GitHub
2 Likes
We have been using Claude to help us develop features in the last year. I didn’t want to hide that fact so I have it sign commits. I fully understand if that’s your dealbreaker. No argument here. For the record, it’s not vibe coded and no features of the software itself use an LLM or are planned to. I’m heavily involved in the whole process, and do extensive manual testing before deploy. The next version would have been out a year ago if I wanted to just push go. I think everything up to Version 0.5, which has most of the core features, was built without agents, though I did use Deepseek for help for some later features. We will never use generated images, and are budgeting to commission some of our favorite Tumblr artists for some upcoming material.
I’m working on a full disclosure of how and why we’ve used Claude, and our plans to stop after the next release. The short version is that, for a variety of reasons, including having Claude paid for and required by my job, a looming sense of burnout on the amount of coding I was doing, and the realities of working in tech at my age, I decided to use Claude to help me sprint to this next version while the cost of these models are artificially low. I see it as a deal with the devil, and I can’t blame anyone for saying “I’d rather not deal with the devil.” But it’s a choice I made with the hope that the tradeoffs will be worth it, that what I can make will be worth the time spent with the devil. I understand that it might not.
The next release, which will be out soon, will do almost everything I and my codeveloper (who is an organic human being and not a chatbot) originally imagined over coffees and notebooks, and it’s built to be usable by people without LLMs, so I’m looking forward to having that out in the world and stepping away from the robot for a while.
I’m glad to be part of a community that has a critical eye on AI, as so many others don’t.
1 Like