Twitter effectively doesn't have an API any more (what they do have is prohibitively expensive for tinkering).
Mastodon has a culture that actively pushes back against many of the things people might want to build - any experiments with things like search or discovery improvements or even potentially algorithmic feeds tend to get a very hostile reaction from long-time Mastodon users.
Bluesky's API is wide open for innovation right now - and you can even connect directly to a WebSocket stream (from JavaScript running on any page) for direct firehose access, like in this demo https://firehose3d.theo.io/ discussed previously: https://news.ycombinator.com/item?id=42159786
Here's my own quick firehose demo: https://tools.simonwillison.net/bluesky-firehose - code here: https://gist.github.com/simonw/af2e50446f6dfc5cb514a7d6aadbb...
I built a small literary journal on early Twitter, and knowing we had a simple way to archive what we were publishing and do stuff like bio-driven contributors notes was a big part of why it made sense to bother investing creative energy on a new platform and social capital convincing writers to make accounts.
It's been frustrating to watch the past few years. I haven't fully decided if I have the energy to reinvent it for a new platform, but poking around at atproto stuff earlier this fall made me more hopeful about the prospect than a similar draft I have on top of gotosocial (even though thw former ia much further from complete).
I noticed the same thing about AI in the Nix community. Was really surprised by the extreme negative reaction, as I could see so many applications of AI that could be useful in the Nix ecosystem.
I did sign up for bluesky, but it has the same failure as twitter: I don't want to hear about your politics just because I follow you.
I wish there was a way to force users to use 1 hashtag per post and then I could unfollow hashtags - but I suspect it would be playing wack-a-mole.
Really anything that aims to replace twitter should be forced to not optimize for engagement.
I built something like this in a slightly different context for myself.
I doubt it'll ever really catch on widely, but I have a bunch of people who want to read my highlights about tech and nothing else who appreciate my software dev feed.[1]
Great thing about Bluesky is that you can easily create your own custom feeds so I guess we can just start doing it and see if it catches on.
1: https://docs.bsky.app/docs/starter-templates/custom-feeds
Although I ended up with a blocklist of around 100 words it was very much a losing battle and a waste of time.
You can still add multiple tags per post, say if it's about music and technology, and if you subscribe to their technology tag, you'll see the post.
Like if you wanted to implement a search engine or recommendation system it’s not clear to me how you retrieve a training set.
Stage 1: new product, hungry competitor wanting eyeballs and users opens up their platform to developers....
Stage 2: developers embrace the attitude and start building amazing things...
Stage 3: the company starts to not like seeing all these products make use of its user base and platform
Stage 4: API features wound back
Stage 5: developers uncermemoniously dumped, API closed
Stage 6: platform hits rocky times, new management arrives, claims to be reborn, reopen platform API to developers
Stage 7: skeptical developers hold back
Stage 8: platform charges $30,000/month for bottom level access
Fool me once, etc etc
> as an open protocol, they literally cannot do that.
From a cursory search, it seems the AT Protocol is solely maintained by Bluesky. Also, AFAICT, Bluesky operates the only sizable/relevant servers. i.e. Neither the governance nor the technical operation are decentralized. (edit: Although the OP call for projects includes options which would probably lead to further decentralization, that doesn't speak to the present state.)
Note that I find it reasonably credible that Bluesky won't do that. I just don't see why they cannot. If there is a massive incentive to make backwards-incompatible protocol changes, say a fundamental security flaw, wouldn't the expected outcome be that Bluesky unilaterally makes those changes?
This is a nightmare for every still independent software/web developer.