• SolNine@lemmy.ml
    link
    fedilink
    English
    arrow-up
    0
    ·
    27 days ago

    I don’t believe there is currently a unified “best practice” of using AI for code development as of yet.

    Context Programming is one avenue that has been used but now we are seeing a lot more product requirement/spec based concepts.

    It doesn’t matter who or what writes the code if it is poorly organized, isn’t tested after each iteration for proper functionality and regression (AI will very frequently cause significantly unintentional regressions), you don’t have clearly defined specs, and your software development foundation is poor.

    Simply telling AI to “do something” often results in it doing it poorly, or with complete lack of context. Develop a plan, tell it exactly what to do, analyze it and review its plan THEN attempt to execute it. Very frequently the core concepts of the plan are incorrect or the suggested fix is incorrect. This is like any other tool, you need to know how to use it or you can severely injure yourself.