@hexaheximal
> Ghost is now slopware

Why do you say that? After shifting the Disintermedia blog to community-hosted Ghost I have a few grizzles, but overall I'm finding it a great improvement on the bloat of SS.

My grizzles with Ghost;

* hardcoded dependence on;

1) commercial email sender

2) one payments processor

* AP module has to be set up separately (how about a single Docker image that deploys both as an option?)

* no footnotes

#newsletters #Ghost

@johnonolan
@FediTips

@hexaheximal
> [Ghost now] Actively built with LLMs

Jesus wept.

Added to the fediverse.party watchlist of apps doing that;

https://codeberg.org/fediverse/fediparty/wiki/watchlist+-+fediverse+apps+created+using+auto-generated+code.-

Please let us know if you learn of any others. For now there's a moratorium on listing any app on the site if the developers merge auto-generated code.

We hadn't got around to listing Ghost, for a few reasons (backlog, and the headscratcher of AP support being a tack-on module), so we won't for now.

#MOLE #AI #VideCoding

@johnonolan @FediTips

@xarvos
> would you count wordpress as fediverse software?

Kind of? We do list their AP plug-in on the dev tools page.

> it has agents.md, has copilot PRs

Jesus wept, another one? Do you know if the AP plugin is itself vibe coded? I guess @pfefferle would be the person to ask about this : D

@hexaheximal @FediTips

@strypey @xarvos @hexaheximal @FediTips only because there is a CLAUDE.md or AGENTS.md does not mean a project is vibe coded.

The ActivityPub plugin has also several agents/skills that ensure code quality, run phpcs or phpunit before code is commited and help with security audits!

@pfefferle @strypey @xarvos @FediTips I can't even begin to state how stupid this is.

If there are LLM agent files, you sure as hell can't trust the project to not commit LLM code. It may not be entirely made with LLMs, but that doesn't change the fact that they took the stance of riding the hype train and therefore cannot be trusted to get off it.

@pfefferle @strypey @xarvos @FediTips And also, even if we ignore everything else: Doing so is still enabling the large-scale exploitation being done by AI companies.

As I've said before: it is outright hypocritical to let the AI companies take your code (especially if it's copyleft!) and then happily use their models without thinking about the implications.

@hexaheximal
> If there are LLM agent files, you sure as hell can't trust the project to not commit LLM code

It's a bit like noticing that a restaurant kitchen has a note on the wall, with instructions for how to add raw sewage to soup. It's not smoking gun evidence that any given soup served in that restaurant contains raw sewage. But you probably wouldn't order the soup all the same ; }

@pfefferle @xarvos @FediTips