user-pic

Dean

Subscribe to feed Recent Actions from Dean

  • Todd Rinaldo commented on 575 Pull Requests in Three Weeks: What Happens When AI Meets CPAN Maintenance

    > In the next iteration of this, may I suggest doing development releases?
    > Let CPAN Testers chew on the code for a bit before you put it into production.

    The issue so far has not been my own module's code it's been downstream deps. CPAN Testers doesn't cover this and definitely doesn't point out incompatibilities with devel versions against downstream modules. So the question I keep asking when people bring this up is:

    How would this help?

  • Todd Rinaldo commented on 575 Pull Requests in Three Weeks: What Happens When AI Meets CPAN Maintenance

    > just need to adjust your AGENTS.md to avoid annoying cliches (at least you didn't say "learnings" or "concrete")

    It's what I get for not being more pedantic with a French person's English pluralisms. :)

    Trust me I actually do correct these several times a week!

  • Grinnz commented on 575 Pull Requests in Three Weeks: What Happens When AI Meets CPAN Maintenance

    I'm not sure where you got this idea. CPAN Testers covers whatever individual testers decide to cover, which certainly includes downstream deps of trial releases. But more importantly: how would it hurt?

  • Tom Wyant commented on 575 Pull Requests in Three Weeks: What Happens When AI Meets CPAN Maintenance

    I don't know how development releases would help you. I do know that just because one of my modules passes all tests on my box does not mean it will pass everywhere. If the problem is truly only downstream failures, maybe development releases wouldn't help.

    But maybe what this means is that we need more testing infrastructure -- something analogous to "Blead Breaks CPAN," but for CPAN itself, not perl.

  • Robert Rothenberg commented on 575 Pull Requests in Three Weeks: What Happens When AI Meets CPAN Maintenance

    I have a lot of concerns about using LLMs.

    The sheer volume of code changes they can submit seems overwhelming. That's a lot to review, and it seems that bugs can slip through. I've seen some daft changes show up in codebases due to AI.

    There has also been some research in poisoning LLMs so that can insert security holes in code, not to mention years of badly-written/insecure code posted online that they have been trained on.

    There are also some serious legal and ethical concerns about using them:

    Do the PRs contain code snippets from other code with incompat…

Subscribe to feed Responses to Comments from Dean

About blogs.perl.org

blogs.perl.org is a common blogging platform for the Perl community. Written in Perl with a graphic design donated by Six Apart, Ltd.