Building Bluesky comments for my blog
natalie.sh
natalie.sh
After spending too much time fiddling with third-party comment systems, I ended up building my own [1]. It's pretty barebones, just does what I need, and nothing more.
Each comment is written to a text file for manual review, so I don't have to worry about spam, cross-site scripting, or irrelevant comments. I usually check them on weekends and add them to my blog.
Comments are stored as plain HTML files, and my static site generator [2] builds the site along with the comment pages [3]. So in a way, it's also a static comment pages generator.
This setup doesn't meet the five attributes (no infra, rich content, real identity, etc.) in the second section of the article, so it wouldn't suit the author's needs, but it has worked quite well for me. I've been using it for at least four years (perhaps much longer, since my old PHP website did something similar), and I've been quite happy with it.
[1] https://github.com/susam/susam.net/blob/main/form.lisp
Personally I find comments not worth the bother and purposely did not include them on my site. My blog is an expression of my personality and the idea of other peoples words appearing on my pages seems weird to me.
I know people enjoy feedback, which is why I have taken to emailing bloggers whose work I enjoy instead of leaving meaningless comments.
So personally I'd be wary of adopting it. I think it's likely the API gets locked down and the comments break in a couple of years.
and would say the bright side of the "enshittification cycle" is that we get nice places for a while and then we can move on. It's not like people party at the Mudd Club or CBGB anymore and why should they? Theory at
They also sold a bunch of shirts as a stunt against Meta earlier this year, and the shirt sales were more revenue than the domains had been.
Data is hosted on PDSs (which are already super cheap to host and are not resource-intensive because they’re just storing data and not doing much computation). They’re not “talking to each other” or serving the web app like Mastodon servers do — they just store your commits like Git and let you subscribe to that stream. I don’t know what “relatively large” means to you as most independent PDS’s are individual people hosting their own data (so they’re actually small) but this sector is growing.
Then, there’s relays (not a lot now but again, they’re now much cheaper to run since Sync 1.1 updates).
And then there’s AppViews (which are more expensive to run because it’s like running an actual app backend with database and all that). Which isn’t too bad if you serve a smaller userbase but is harder if you want to serve millions of people like Bluesky’s own AppView does. Normal considerations of “running a web app for millions of users” apply here. But keep in mind that people on all backends see exactly the same data — it’s not being sent back and forth like messages between “federated servers” like on Mastodon. It’s being aggregated from the entire network via relay(s). So if I want to run a Bluesky backend (aka AppView) for a small group of people, I can do that relatively cheap and see the same content as the rest of the network sees. If my content goes viral, my servers won’t go down because going viral has nothing to do with the load on AppView. The load on AppView is just about how many people are hitting my own backend from their browsers (apps) — nothing to do with which server I have chosen to store data on (PDS) or to host my custom web app on (AppView).
So when you say “list of servers”, what kind of servers are you referring to? They all have different incentives to run (I’ve enumerated them above) and different relation to the user (PDS is something you choose, akin to hosting; AppView is determined by your client; relay is kind of behind the scenes and determined by AppView). And if you’re concerned about the load, I want to better understand which scenario you’re describing.
You aren't wrong, there will be a turn at some point just like there was with Twitter, but then the same is true of 'AI' and people seem happy to go all-in on that. If VC's want to burn their money on the dream of becoming some new kind of rich... good, let them. Sure it turns sour after a while (Uber, Doordash, etc)... but enjoy the largess before they figure out there's no magic money tree in those hills.
...and you loose access to all your messages and network.
I'm currently building a review system for my open source Web map https://cartes.app, based on Bluesky. Not trivial though, you have to create a lexicon and maintain a DB based on the Bluesky stream.
2 - https://docs.bsky.app/docs/api/com-atproto-repo-get-record
Bluesky _will_ enshittify sooner or later
Better go with my own DB.
Or use a network with a well-designed protocol, a hosted service, 30 million users, a social graph, moderation...
Many OSM apps will also be reluctant to adopt a closed source solution that might be closed of any moment. And under what licenses will those reviews be? As MapComplete developer, I can not and will not be adopting a system based on Bluesky
Bluesky is not closed-source, it's mostly open. It's a hard mastodon-circle myth. Aside from being open-source, it's a protocol.
What I'm doing with Bluesky should be easy to reproduce with Mastodon, opening to 10 M accounts more.
People won't create an account to review. Not even a OSM account. We need to build on what's available, and ATProto is spot-on for this usage.
Thousands of people have already set up their own PDSes and it's inevitable managed hosts will appear soon. Blacksky just started migrating people over to its own PDS. AT's credible exit is close to reality after about two years while all the promise of ActivityPub and predecessor protocols has yet to materialize after over 15 years.
I only see 2 posts on the entire blog, both from 2025 (and one is this post).
BS is an attempt to recreate an even more toxic environment than old Twitter ever was.
Which is all very high school cafeteria-drama.
The various blocklists are opt-in; you’ll only be invisible to their respective subscribers. Only the default bluesky moderation list is global, and they only adjudicate ToS violations (like every other social network).
Community moderation is quite distributed and egalitarian on bsky, perhaps even more so than the benevolent dictatorship used here (which obviously doesn’t scale).
> BS is an attempt to recreate an even more toxic environment than old Twitter ever was.
On Bsky I have yet to have anyone out of the blue, with no prior interaction, call me a slur or racial epithet. Can’t say the same about my old Twitter account.
And of course it's also opt in as well. Just the default bluesky client does that by default. Any third party client (ex: https://deer.social or https://zeppelin.social) can opt-out of "default moderation". And technically you could use a userscript or even potentially a ublock rule/filter to disable default moderation (just like you can to disable regional moderation or age verification).
This means it's opt-out. Not opt-in.
The reason I said it's opt-in is because moderation is added client-side by including the moderation service's DID in the `atproto-accept-labelers` HTTP header when sending requests to the appview.
So it is by-design opt-in, just in practice the "first party bluesky client" makes the choice for you for legal compliance reasons, and with an increasing hint-hint-nudge-nudge from the devs to use third party or forked clients to bypass the various legal restrictions countries keep trying to impose on them.
Seems to me like people who subscribe to a blocklist that I'm on aren't people I want to be visible to/communicate with.
More people have HN accounts than bluesky accounts. Bluesky is pretending to be a public forum, while HN is purely proprietary and closed solution - I'd much prefer to depend on bluesky than on HN (despite already having an account). Also there are limits to acceptable self-promotion here.
>As for rich content, words are enough IMHO.
As you correctly noted, this is your opinion, that the author of TFA doesn't share.
You really think that?
it has guest support (so people does not need a matrix account to comment), but if you use your own matrix account, you are essentially joining a matrix room per post.
Over the years I've found that any interaction with social media that involves me loading the feed inevitably ends up with me doomscrolling or spending way too much time scrolling stuff that doesn't add anything to my life. This could be a way to avoid that cycle, finally, but still interact with the wider social media world a little bit.
Almost as if it was designed that way...
If BlueSky banned you tomorrow what is the plan? If BlueSky went bankrupt tomorrow?
I figure there are other AT compliant products that you can switch to but a lot of data would go missing?
Is it perhaps that the "right-leaning" social media users being banned are also violating the rules? Attacks and abuse seem to be standard practice, especially for the American right.
Please don't break the site guidelines like this, no matter how incorrect another comment is or you feel it is.
Doing this has the obvious downside of making the threads more toxic, plus the less obvious one of discrediting the truth (assuming your comment is indeed correct) by giving it toxic associations. https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...
One can't conclude that for sure, it's unnecessarily personal, and it doesn't add anything (other than a swipe) to the correct information in your post.
And PDS level/"account" bans are just at the PDS. If you've been "banned", that's just bluesky the PDS host telling you they don't want to host you and that you need to go host your data yourself or find someone else to host it for you. i.e. find another PDS.
Basically every form of ban or moderation in atproto/bluesky is "soft" moderation and you can fairly trivially bypass it and continue doing your own thing.
The overwhelming majority of right wing accounts that get banned do so soon after joining (and generally after going to pick fights). And they never even bother to try to keep their accounts, instead choosing to create new accounts to get banned again or abandon the platform. It's disingenuous behavior and for right wing personalities it feels almost more like a sticker of pride that they were "banned from bluesky".
Plenty of right leaning and libertarian accounts exist on bluesky. Project Liberal [1] and Liberal Party USA [2] (run by Josh Eakle[3] and Kevin Gaughen[4] respectively) exist just fine on bluesky and they are large splinter groups from the Libertarian Party following the whole Mises Caucus coup attempt thing. Likewise a number of libertarian groups such as the Libertarian Party of Lousiana [5] do just fine on bluesky. And of course AI and Cryptocurrency people also do just fine on bluesky as well despite the stereotypes against them and the common belief that "they aren't welcome on bluesky". The worst thing that happens is people block or mute you and you don't have to deal with them anymore rather than toxicly fighting each other each time you see each other.
TLDR: Everyone is welcome on bluesky but there's no requirement for people to tolerate you. Even if you violate every transgression, as long as you aren't posting literal child porn to the network you'll still be able to exist just fine however people might just ignore you.
------------
1. Project Liberal: https://bsky.app/profile/did:plc:d5nigw7kzpsglf3gtl2dvbev
2. Liberal Party USA: https://bsky.app/profile/did:plc:v3jmda7lwwdoofcvgmjwsbcg
3. Josh Eakle: https://bsky.app/profile/did:plc:2adtngm3y6e6ol6jastnkxzm
4. Kevin Gaughen: https://bsky.app/profile/did:plc:4oyecf2hz4ajhm4zqp52hxqo
5. LP of Lousiana: https://bsky.app/profile/did:plc:mongiv55fh5l5e7vi7cbjajw
LPLouisiana is definitely left leaning libertarian but the first 4 are all very much your old school small government libertarians.
Both Josh Eakle and Kevin Gaughen used to be senior members of the Libertarian Party prior to the Mises Caucus burning it to the ground and they are absolutely center right libertarians through and through.
No matter how many times this pedantry gets repeated in this thread, this is literally not opt-in.
A light switch that is glued in place so you cannot turn it off is not opt-in.
Sure, some people with the know-how can get a pair of dikes and cut it out of the wall, the light will turn off and they will say "see, I opted out!"
But most people won't do this. At best it is misleading to say so.
The goal is that third party apps take over as the majority share of clients over time and the main impl should be seen as a "reference implementation".
And that of course ignores all the other non-bluesky projects currently incubating on atproto.
"do just fine" is really pushing it I think. I don't think the core mod team or the default-client's moderation service is biased toward the right or the left (at least in how they apply moderation, not necessarily their personal beliefs.) I think the core team is doing the best job they can given their resourcing. The community on the site is a different story.
There was an attempt by AI researchers to join the site and they all got bullied until they left, largely by the community. Pretty much every reply to an NYT article that doesn't denigrate Trump is either "wow how does the NYT have the time to write about fashion/lifestyle/<anything but how Trump is awful>, it's because they've an evil right wing publication" or "how dare the NYT platform this opinion it's an evil right wing opinion". If you look at feature rollouts or social posts by the team you get lots of well-liked comments about how the mod team enables right wing behavior.
I use the platform a lot and am really rooting for them to succeed, but I feel that there's just a lot of lefty toxicity on the platform and that the community on there loves politics and often brings politics into unrelated threads. They seem interested in a sort of pop politics too, not the kind of harder political analysis that a good think tank or non-fiction book can provide either. I feel that if you're into lots of pop left politics and the culture that emerges from a community with this love then you'll like Bluesky. For now I think the community is too political to really foster wider conversations the way pre-Elon Twitter did.
The vast majority of people using Bluesky don't use a 3rd party or self-hosted PDS because they aren't and probably shouldn't have to be aware of it. It isn't even a prominent feature. At least with Mastodon, it's a big part of the culture at its' core... you don't see this with the official app even.
I didn't even know that was even an option until you mentioned it. I only know that there have been banned accounts on the official/default one that didn't even post anything yet.
Rather, there is an association between modern conservative/libertarian voices and populist messaging - and all the pitfalls that come with it. Meaning, vulgarity, emotional bait, deception, and purposefully offensive language.
Like, the modern American conservative leadership cannot advertise their own ideology without resorting to lies and attacks of character. The left, by comparison, does not operate that way.
So, if you're censoring shitheads who are generally unliked, that might appear as though youre targeting conservatives or right leaning people. But you're not.
Basically, the right has purposefully positioned themselves to be associated with unpalatable ideas in order to leverage populist messaging. And this worked - they won an election. The downside is that now if you filter out unpalatable ideas such as blatant slurs you're going to necessarily mostly target right wing people. By accident.
(blog author works at bluesky, no affiliation personally)
And of course there are several implementations and hosts for relays (the gossip nodes), PDS implementations, clients, and appviews (the server backend for bluesky the web app).
So strictly speaking if bluesky imploded tomorrow you could just use a self hosted version of the same app or use someone else's (such as https://zeppelin.social).
The PLC directory is still technically in bluesky's hands but is being transferred an independent foundation atm and could be trivially forked if needed. And of course if you use did:web that doesn't apply to you and you just depend on DNS.
So if there's a large scale exodus from bluesky, as long as full backfills of the network exist, you'll be able to reconstruct your CAR files, etc even if your PDS dies.
So yes if they die all the comments disappear but people can reconstruct their history and move it to other PDS like blacksky, northsky, or others who are getting ready to start onboarding/open enrollment.
TLDR it'd be a bit rough if they died overnight but if it was a slow death and people had a bit of warning you'd see people move on to other PDS without issue.
However, atproto data is append-only and cache-friendly. It wouldn't be hard to record historical comments and join them to the ones returned by the live query. (I'm probably just going to script periodic backups for mine and worry about displaying them when/if BlueSky does dissappear.)
It’s a completely valid reason, but All the talk about platform lock in, independent nodes and relay and whatnot is just to make you feel better (I listened to some talks and podcasts but realized that it’s all just window dressing, you can be practically deplatformed at any time, so I’m hazy on the details).
Credible exit matters too. The idea that all of this data is signed, public, and verifiable by anyone helps ensure that the provider is accountable to the community. A provider might decide against enshittifying a product if they're worried about a competitor doing better and taking their place.
Yes, an atproto lexicon will tend to be dominated by a small number of nodes. It's not the perfect distributed system some would prefer, but it is a real improvement on the completely centralized models seen everywhere today.
- Run the comments on an instance you moderate
- Even better, only show comments that your account has favourited.
More details on the last one here:
Yeah I like this solution. Might try explore this approach
If I only show comments from a well-moderated Mastodon instance that my main account is on, plus any instances that it federates with, does that not solve the problem?
Presumably the blog interface itself can choose to simply not surface hidden replies at all; if you view the thread via a different client (eg the Bluesky app) you would have the option of seeing the hidden posts.
And of course if you view the thread through your own Bluesky interface your personal blocklists and moderation would apply to the thread.
I built a web component for the same purpose, and you can see in there how I implemented threadgating: https://github.com/ascorbic/bluesky-comments-tag
Always looking for new places.
https://github.com/AndreasThinks/quarto-open-social-comments
https://cassidyjames.com/blog/fediverse-blog-comments-mastod...
Neat integration with bluesky, though.
Security's better now in general. But the same hands-on approach sometimes brought us widely exploited 0-days. Drupal was obnoxious. Yet people knew throbbing banner ads were cross-site things in a way you could turn off.
As for frontends, there's a bunch of them and a lot of them focus on changing the UX. But for self hosted "bluesky", there is https://deer.social which is a forked client that still relies on the bluesky appview/backend and there is https://zeppelin.social which is downstream of deer social but also runs their own appview independent of "big bluesky".
One example of a profile that lives on a different PDS ("AT Protocol Personal Data Server") than the default one: https://bsky.app/profile/mackuba.eu
If you look in the bottom left of https://clearsky.app/mackuba.eu/history, it shows:
> User changed [...] 1/2/2025 to external server PDS: https://lab.martianbase.net
If you go to that URL, you'll see the landing page for the atproto PDS software.
Edit: Ah, and seems Kuba is hosting a directory of profiles using their own PDS as well, lots of examples over there: https://blue.mackuba.eu/directory/pdses
Biggest one right now (excluding the default one) seems to be atproto.brid.gy, which has 40393 accounts.
1. If you switch PDS all links continue working.
2. If you change your handle (for did:plc, did:web can't do this because DNS) it used to break links but nowadays this isn't a problem because handle resolution respects historical handle naming (I think it works by post+handle age but I can't remember).
3. Also if you share posts using the did syntax instead of handle syntax (which bluesky seems to be slowly changing over to, at least profiles do this now), it's stable regardless of handle changes.
4. If you want to switch frontends, you can use an extension or app like at://wormhole to do so. UX for this should improve over time but that's a big "eventually".
5. Hopefully the at:// URI format catches on but that's a long ways away given that browsers make using custom URIs an absolute nightmare.
The default Bluesky frontend uses bsky dot app URL when you use the "Copy link to post". Now if one day, you lose trust in this server and switch your PDS, this link continue working depends on this very non-trusted server. If this server is profit-seeking, it can break such links.
An extension or another app is not the solution, and neither is the new at:// URI format, because what matters is the relationship the default server sets up with its majority of users. Most bsky users will lose their traffic to their own posts and therefore will be locked-in, cementing this one server to be the dominant one in all perpetuity. We will therefore get all the patterns of monopolistic abuse that we have seen elsewhere.
I mean it’s fine, use whatever your comfortable with and Bluesky is the next frontier for development ideas.
EDIT: Ask five people you know outside of tech if they have a Github account. Everyone I know outside of tech moved to Bluesky from Twitter. No one I know outside of tech has a Github account. If I encounter someone who has neither, I'm of course going to recommend a Bluesky account from a utility perspective, as they're likely never going to contribute code, issues, discussion on GH if not a tech person. (most of my network is non tech, non startup, non SV people, ymmv; HN is the closest I get to tech folks most of the time)
And even then while "daily active likers" is down half from 6 months ago, it's still up substantially from even just a month or two prior to that. Bluesky exploded in size at the start of the year and it seems to be finally settling into a steady state with gradual growth (vs the prior explosive growth + falloff).
"Daily records" (at the bottom of the page) is a bit better metric of overall network activity and even though it has also fallen since peak it shows there's still an order of magnitude more activity on network than prior to blowing up.
Bill Corbett: https://bsky.app/profile/billcorbett.bsky.social
John Scalzi: https://bsky.app/profile/scalzi.com
Wil Wheaton: https://bsky.app/profile/did:plc:mwpiq2rr6joccohcz2urkwvn
NY Times Pitchbot: https://bsky.app/profile/nytpitchbot.bsky.social
Electrek: https://bsky.app/profile/electrek.co
Stephen King: https://bsky.app/profile/stephenking.bsky.social
I’m among that 31M who signed up (as are many of my friends) and only the most left ones are still using it. I trolled for a couple of days until it got boring.
I would be willing to guess that self hosting gitea as a backup mirror is less work than doing the same for Bluesky. But, just speculating