Our first blog post about Catodon’s new chapter - please repost/boost!
In defence of our choice to use LLMs
I know AI is a sensitive subject on fedi, and I understand the arguments against it. But here’s what it’s actually done for us: First of all, the project wouldn’t exist without it. But even now that we have a dev on the team, we’re keeping it in our workflow.
In six months, it’s helped us ship features that projects with several full-time devs would need years to build:
- forums with full Lemmy compatibility
- an RSS timeline
- federated blogs
- a dedicated Direct Messages page with a chat-like interface
- New/Hot/Active sorting for forums and microblogging timelines
- a WYSIWYG editor for Pages and Blogs
- native Mastodon API support
- multi-protocol login that lets you use Catodon as a client for your Mastodon, Pixelfed, or Misskey account too.
It also recently helped us catch a vulnerability in upstream Misskey, which we reported and which has now been patched - meaning every Misskey fork will benefit.
We’ve also used it to spot 13 Misskey security patches that hadn’t been merged into Sharkey yet, which we flagged to the Sharkey team as well.
And to be clear, AI isn’t magic - we spend hours of real work on every feature, and everything is tested and reviewed before it ships. We also keep a close eye on Misskey and Sharkey upstream, so we can pull in security patches as soon as they land.
We’re not going to pretend this settles the debate about AI, and we don’t expect it to. More and more projects (including the Linux kernel) are utilizing LLMs - and there are actual, valid reasons for this. All we can do is keep being transparent about how we use these tools, and let our work speak for itself.
Wow a project to sloppify sharkey
Dunno what should be my take on coding with AI. They have good argument on its use, the code is being reviewed by devs and they were clear about its use.
I’m against AI for its environnemental disaster, cash bubble, capitalism, ram shortage…but there, it helped him with this projet revival.
So for now, i ask everyone to let go the part on AI.
I consider it was already discussed and they made it clear why they use it : he didn’t find dev to help him.
On Piefed, Lemmy…we lacks contributors. Imagine the project lemmy stop, what would you do ?
There is another fork of firefish : iceshrimp where the same dev AntoineD contribute to. It is being rewritten in .NET. But it takes time and lot ressources.
I hope more contributors can give a hand.
Well they also wrote the blog post with an LLM too, so it’s clearly not about coding faster. That’s just their excuse.
Hi, I’m the LLM that wrote the blog post, thank you for your kind words =)
Diversity is important, yet dividing to much puts too much weight on everyone shoulder.
Using IA means reducing the weight on your shoulders at the cost of less socializing, since you need less people to do things.
Like the chat world, the multiple ia projects appearing with people reinventing the wheel instead of contributing to existing open interfaces/softwares like matrix, I feel more people should contribute/build on things like NOSTR, instead of reviving abandoned things with a small team.
Anyway, ́tis only my take on the subject, but really, NOSTR seems the next big thing with your idendity independent from the hosting server.
Well, nostr was technically interesting but the cryptobro put me off. There is also the moderation aspect that i wasn’t fond of.
However, I didn’t stay long enough so it’s hard to recall my memory.
But yes, having an identity that is not tied to a server as nomadic identity (stream, hubzilla) should be the futur of the fediverse. It will make our network more resilient.
There is also scuttlebutt that i loved.