This appeared in my feed today and I like the idea of it.
Bubbles seems to be a “Hacker News but for non techy blogs”, that curated a bunch of RSS feeds from blogs and every post is shared, but some float to the top (the ones that have votes), and some sink down.
Nice analogy, hopefully it ends up a successful project 
So fun! Thanks for sharing this.
My own blog has appeared on their list as well, haha.
It sent me to an awesome blog post right away, and a few more once after looking around a bit. Good to see some familiar names on the front page, too. This one’s a keeper!
This was also shared in the Discord! I’m just going to copy/paste my thoughts because they haven’t really changed.
Hacker News and Lobste.rs have community voting figured out, but non-tech content gets drowned by the tech majority. Kagi Small Web curates thousands of personal sites, but has no community-driven ranking. Blog directories help you find blogs, not today’s best blog post. Social platforms own the conversation. Mastodon is decentralized and ad-free, but you only see what the people you follow share. RSS is great, but solitary. There’s no collective signal telling you what’s worth reading today.
Not sure if I agree with this 100%—I’m most certainly missing out on feeds I could enjoy, but I do easily find feeds I do. It’s fine for things to slip through the cracks. It’s fine, IMO, for people to have to do the hard part and search for stuff. It’s how you ensure that the feed really is something you’d like rather than something you’re iffy on but mindlessly follow anyways.
Also, depending on the culture/makeup of the community, community voting could easily have similar problems as an algorithm. While algorithms trend one way or another based on unseen code, their main direction—at the end of the day—is still fueled by how people interact with it. Even if some algorithms do a worse job than others at guessing what people really want lol (or if some are particularly programmed to show/not show certain content).
This does still seem intresting if nothing else but… I’m still a little weary. Especially since they seem to oversell the idea of how hard feed finding is without Bubbles, to make Bubbles more appealing, instead of just… letting it stand on its own merit.
Maybe it’s just a creator being proud of their latest project. Besides, it matters who’s doing the curation. There’s ooh.directory for example, but I never, ever found a good post through them, and stopped visiting after a while. So having choices is important.
I think this person is coming from a place of good faith and wanted to make a project for themselves that they’re also sharing for others to participate in.
I don’t think that they mean for this to be a replacement for the other ways to discover blogs, but just another option that has a similar system to Hacker News and Lobste.rs.
edit: As someone that posts a lot of life / non-tech content and browse through the bear discover feed I do appreciate seeing a similar type of feed that is geared towards non-tech things. (Though I still mainly use my own RSS feed when I follow blogs).
I guess it’s just the phrasing of “telling you what’s worth reading today” that sets me off. My critisms are not an attack on the project by any means — just a genuine concern that it’s re-introducing ideas to RSS that many of us (myself included) are explicitly trying to avoid when using RSS. I don’t like the idea of feeds becoming popularity contests, posts being tailor-made to the desires of this community (or similar ones) to make Number Go Up.
the layout reminds me of hacker news and lobste.rs also, bearblog has this thing that lets you vote on blog posts
That does seem to match what’s in the FAQ, where it says Bubbles is for “People who enjoy reading blog posts about things other than startups and JavaScript frameworks.”
Since that’s… most blogposts, this strikes me as a project that could stand to narrow its scope. I’m reminded of the times when I’ve prompted my students to identify a target audience and they simply responded “the general public.”
I share the same sentiment. If someone’s telling me they can identify “today’s best blog post” on my behalf, I want them to tell me their criteria. If their criteria involve Reddit-style popularity contests then I expect we have different priorities.
I also follow dozens of hand-picked RSS feeds. An aggregator is complementary to that. It can surface blogs and posts I would never had thought of searching for by myself.
Oh look, another good one this morning. Going to keep using this service for a while.
Thank you for sharing! Looks like they do source from there and others FAQ — Bubbles. I was curious how they get the sites in there, submission or otherwise. I set up my robots.txt to not be scraped by AI bots, but if it’s via the Indieweb ring or Bearblog I’m fine with that since I opted in to be there.
Update: turns out Bubbles.town is AI slop. And to think I defended them. This is why we can’t have nice things.
Aw, crap. That sucks. Though, reading through the FAQ, it did have some LLM stink on it. And I was already suspicious because the project isn’t open source.
It was a really cool idea, but I’ve just emailed to have my website removed.
Thanks for pointing that out. Copying this over from the FAQ (edit: sorry, wrong link, should have said Privacy page):
AI classification
Bubbles uses Anthropic’s Claude Haiku model to assign categories (for example Tech, Science, Nature, Writing) to posts and blogs. This classification is purely organisational: it decides which category a post appears under on bubbles.town. Nothing more.
The data sent to Anthropic is limited to content the blog owner has themselves published for syndication, specifically: the blog’s name and the titles and excerpts of its public RSS or Atom feed.
So in other words, Bubbles uses a large language model to assign category labels, and it also sends everything in your RSS feed to Anthropic.
Evidently this came up when another blogger, unaware, praised Ben for carefully vetting thousands of blogs, and then, in the spirit of supporting independent developers, rewarded him with a donation.
If Ben has any integrity then I think that donation should be refunded.
Woah thanks for surfacing that Coyote! I don’t agree with the developer that maintaining an RSS feed is the same as consenting to LLMs processing that text. I’m also wary of aggregators being opt-out by default instead of opt-in.
At least Reddit and Hacker News require a human to submit a post. It’s very different to say “Hey, I will share this link because I found it interesting” than to have scrapers hitting servers across indie sites (some self-hosted or running on infrastructure not expecting large volumes of traffic) to gather content.
On the one hand, from what I can tell, after the initial application of categories, Bubbles itself only draws on the RSS feed at the normal rate of any RSS feed reader. On the other hand, the initial assignment of categories has been relegated to a type of program which has notoriously gorged itself on continuous data scraping.
Anyway, something I should have aired my thoughts about the first time around:
Part of why I balked at this project to begin with, even before I knew about the LLM factor, is the framing here as if blogs are very hard to find and as if you can’t be expected to discover things without this kind of mass automation.
From my perspective, finding even one blog should open the doorway to many others, and if you’re having trouble traversing the blogosphere then that tells me you’re looking at blogs that don’t use links.
Then again, perhaps traversing the blogosphere is not the goal of the kind of person who consults an LLM to assign blog categories and who wants to rely on Reddit-style popularity contests to determine which post is “today’s best.”
@nosycat don’t feel bad. I was looking through this thread and agreeing with you, so my first instincts were the same.
I wouldn’t say it’s “AI slop” because to me that term means low-quality content generated by AI and the content on there are largely other people’s blog posts.
I find it interesting that the site notes that AI-based blogs are not allowed… but then uses AI to categorise posts. I feel like a better approach would be to categorise blogs by hand to make those “subfeeds”. Would that mean fewer blogs? Probably, but it would certainly be nicer for curation. I’ve also noticed that this isn’t a directory for manually-submitted blogs or the creator’s picks: it’s automatically gathered from various sources. The problem is that at least of some said sources already have discovery systems in place. Bear’s discovery feed shows you new and popular posts. They are not sorted into categories but that just takes us back to the AI issue.
The owner says they respect rules in people’s robots.txt and that they can disallow this particular bot but I feel like making this opt-out rather than opt-in a terrible choice, since many people might not want their posts fed to Claude. I disallow all bots and prefer to manually handpick which ones to allow, but I know many people don’t want to have blanket rules like that due to legitimate bots they might be comfortablr allowing. This just kind of ruins that, in my opinion.
I love discovering blogs so an alternative to oooh seemed nice… but not like this :(
@Coyote bearblog has an optional “vote” feature. I initially removed that button from my blog but ended up allowing it. I’ve enjoyed many of the posts that end up “popular” on their bear. It gives me an idea of what other people are reading that isn’t just the blogs I follow, and the blogs my favourite bloggers follow, and so on. I hate most “new web” features but I won’t throw the baby out with the bathwater and say that all voting features are bad and a popularity contest. It really depends on the community, in my opinion.
That made me laugh out loud 
On the topic of Bubbles, I’m indifferent at best. These sort of communities always end up as popularity contests. Great if you’re writing for exposure and clicks! But generally antithetical to what I consider the main point of the small or indie web, or whatever you call it. Which is to discover all sorts of interesting people and esoteric content. And I don’t think that’s achieved by a voting mechanism.
But I applaud the project nevertheless! Anything that creates attention around the independent web is a good thing.
The Bearblog upvote feature seems to be basically a “like” button and the only direct on-site way to interact with a post at all, so I can understand why you would put it back.
From what I can tell, the Bearblog ecosystem benefits from being relatively niche, meaning the userbase largely has something in common (via “being the kind of person to use Bearblog”) to begin with. This may give its popular feed a relatively consistent feel, even if it’s not explicitly themed.
With its upvote system elevating posts that get a lot of votes, though, it seems rather vulnerable to the possibility of, hypothetically, for instance, an outside demographic coming in with otherwise unpopular views and blasting up to a high number of votes from outside visitors, thereby inflicting that kind of thing on the main userbase without much option to filter it out. And by “hypothetically” I mean I have already seen people reacting to exactly that.
Point being, while Bearblog is a relatively benign example, I think it also illustrates some of the potential pitfalls.
Anyway, back to Bubbles, if there’s no explicit theme and the determination for “best” is being left to the userbase… I’d want to get a better read on what to expect from the userbase. At the moment, all I really have to go on is that Bubbles requires a Fediverse account (which excludes me). The userbase may or may not also lean toward people who are okay with involving the Anthropic bot, but I don’t know yet how far the news of that has spread.