Just a PSA.

See this thread

Sorry to link to Reddit, but not only is the dev sloppily using using Claude to do something like 20k line PRs, but they are completely crashing out, banning people from the Discord (actually I think they wiped everything from Discord now), and accusing people forking their code of theft.

It’s a bummer because the app was pretty good… thankfully Calibre-web and Kavita still exist.

    • Jolteon@lemmy.zipEnglish
      211·
      3 days ago

      It seems like the criteria for making it on there is fairly lax. Nextcloud makes a list by simply having an AI assistant as an optional (user-facing) feature, while none of the actual code appears to be AI-generated.

      • Evotech@lemmy.worldEnglish
        12·
        3 days ago

        I’d be easier to make a list of all software that doesn’t use any AI at this point

      • lambalicious@lemmy.sdf.orgEnglish
        71·
        3 days ago

        Even if the “has an optional AI assistant” was not a thing, the repo includes an AGENTS.md file, which is also listed in the criteria, and more than qualifies it as slopware.

    • nfreak@lemmy.mlEnglish
      332·
      3 days ago

      Man this list is depressing. Good to have handy though. Sad to see SearXNG and a few others on here.

        • Auli@lemmy.caEnglish
          182·
          3 days ago

          Did you read about kernel they are experimenting with using it for reviews. They have some prompts for LLM to catch issues before it gets to maintainers so it frees up time. Don’t see an issue if that is all it is.

          • lambalicious@lemmy.sdf.orgEnglish
            92·
            3 days ago

            It might be, but for some people that might, understandably, be already bad enough, a line in the sand if you will.

            I’m reminded of this statement about LLMs and the kind of people who use them in the first place. It’s an early indicator that quality (and sovereignty) of the software is going to go the incline down.

          • PoliteDudeInTheMood@lemmy.caEnglish
            31·
            2 days ago

            It’s not, the second I cloned it and gave codex access it found a whole whack of privacy issues. This was 100% human coded

            • fccview@lemmy.worldEnglish
              31·
              20 hours ago

              degoog Dev here, definitely not vibecoded. Would you be able to tell me all these whack of privacy issues? I thought I had everything covered, but if you found something concerning it’d be nice to know before I get it out of beta :)

              • PoliteDudeInTheMood@lemmy.caEnglish
                4·
                11 hours ago
                1. Fixed credential-exfiltration risk in /api/proxy/image: Previously the endpoint could:
                • accept arbitrary auth_id
                • load stored API keys
                • forward them to attacker-controlled URLs
                1. Enforced outbound host allowlist globally Previously:
                • allowlist existed
                • but outgoingFetch() didn’t enforce it
                • plugins/engines could bypass it
                1. Fixed extension store path traversal Previously a malicious store manifest could:
                • inject … paths
                • escape install directories
                • reference arbitrary files
                1. Hardened proxy IP trust Previously:
                • rate limiting trusted any X-Forwarded-For header
                • clients could spoof their IP
                1. Fixed inconsistent settings authentication Previously:
                • settings UI stored an auth token
                • but the settings modal didn’t send it when saving
                1. Implemented Improved proxy deployment support
                • Added proxy-aware behavior:
                • DEGOOG_PUBLIC_BASE_URL for canonical URLs
                • secure cookie handling when X-Forwarded-Proto=https

                Additional Improvements:

                • suggestion fetching hardened
                • DuckDuckGo suggestion parsing fixed
                • unified outbound request handling
                • install state guard properly cleaned up

                Made some other changes for my specific deployment. Very happy with your work so far. Thanks so much

                • fccview@lemmy.worldEnglish
                  2·
                  10 hours ago

                  Thanks, I’ll individually look into all of these ♥️ I’ll say some of them are more conscious compromises for the sake of an open scalable system where third party extensions can truly edit anything (intentionally) and everything around Auth/secure cookie is also fairly lax due to the fact the Auth is just a protection for the settings (which literally stop the settings from being served by the client), in the moment I decide to add some more structured Auth system/maybe users I’ll look into proper secure cookie handling.

                  This is an awesome report, thank you so much for sharing it!!!

    • Sips'@slrpnk.netEnglish
      83·
      3 days ago

      Holy shit I almost whish I didn’t read through that list…So sad to see so many projects go down this path…

      • lambalicious@lemmy.sdf.orgEnglish
        2·
        3 days ago

        Geez… problems never end, do they.

        I’m barely active in Codeberg. Unless someone beats me by, say, end-of-month, I might file an issue about it; that said, I’d like to be able to offer at least one (1) functional alternative rather than simply +1’ing to the complains that this or that is Never Good Enough.

    • uzay@infosec.pubEnglish
      2·
      3 days ago

      Especially since it is actually listed as an alternative to calibre there