Where Fluxer is headed in 2026 and what we’re working on next.

  • paydequeso@piefed.social
    link
    fedilink
    English
    arrow-up
    6
    ·
    8 days ago

    They do have an LLM usage policy, if anyone is interested: https://github.com/fluxerapp/fluxer/blob/main/.github/LLM_USAGE_POLICY.md

    The tl;dr is: you can use an LLM for private research, but you cannot use an LLM to write anything (except translations).

    This line is kinda sus though:

    This policy applies to every person who does not have write access to this repository.

    … So… is this Hampus carving out an exception for himself? <x-files.mp3>

    Overall, I’m a huge Fluxer fan. I think they’re doing the right things 90% of the time. I’ve never seen them enthusiastically embrace AI, but they haven’t been vocally anti-AI either. The LLM policy does basically seem like a ban though?

    IDK. I suppose the large amounts of code churn may be a sign of AI? That’s definitely not conclusive though. Any software engineer will tell you running a code generator (not LLM, we have to regenerate our protobufs every time we change our API) or a code formatter or static analyzer (updates a bunch of files) or vendoring your deps (shout out to Go devs) will also inflate your line count.

    I also checked out a few random files and they don’t look like the slop I’m forced to generate at work. I glanced at like 5 files. I guess Hampus could be tuning Claude to hide the AI slop better? IDK… this is all starting to feel like a conspiracy theory…