• jobbies@lemmy.zip
    link
    fedilink
    arrow-up
    3
    arrow-down
    1
    ·
    9 hours ago

    You use them minimally and responsibly, you don’t over-rely on them. Is there a problem in that case?

    Responsibly is the key word there. I.e. vibe-coding an entire project full of security risks without understanding the architecture or languages used. Vs. using LLMs to error check code, brainstorm a project, write code you feel confident you can review yourself.

    Also, like you say, hosting locally. Or using a provider like Mistral that is European, not evil, uses France’s abundant nuclear power…

      • jobbies@lemmy.zip
        link
        fedilink
        arrow-up
        6
        ·
        8 hours ago

        I sort-of agree, but what are you disclosing?

        If a project on github was 100% vibe coded with no (initial) human review - I’d want to know that.

        If a project was 60% hand coded, 40% vibes but reviewed by a human … yeah in the interests of openness it doesn’t hurt to disclose that.

        If a project was 100% hand cranked but checked by an LLM? If its open source I’m not sure I need to know that.

        If a project was 100% hand cranked and human-reviewed, is open source but the dev used an LLM to keep a to do list of tasks… no I don’t need to know that.

        And like someone else pointed out - how exactly do you verify that AI was not used at all for any part of a project?

        Also, you know Microsoft will be vibe-coding its products. Would that stop you using Windows, Xbox, M365 etc?