855
Amazon AI (thelemmy.club)
submitted 7 months ago by to c/programmer_humor@programming.dev
81 comments fedilink hide all child comments
 
you are viewing a single comment's thread
view the rest of the comments
[–] 322 points 7 months ago* (35 children)

People who share the size of a codechange as a mark of how effective ai coding agents are truly missing the point of code changes.

  • permalink
  • fedilink
  • source
  • hideshow 35 child comments
    •
  • [–] 123 points 7 months ago (18 children)

    Number of lines of code written is a shitty metric to measure productivity.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 18 child comments
    •
  • [–] 55 points 7 months ago (1 child)

    And yet I experience it so often. That or “effort points” as the metric being used to determine who all stars are.

    Either as a metric just encourages gaming of the system:

    • Why write one line when I can write the same thing in 20?
    • Why take this one effort point task I think will take three when I can just skip it and grab these one effort points I think will take 20 minutes?

    I’ve been on teams that on the surface didn’t have these metrics matter, but the top effort points achiever got bonuses on the DL.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 29 points 7 months ago

    What did you do?? You refacted the code and now it's better organized but you overall got rid of lines?

    I'll set up a PMD meeting to help you out of this problem, but fair to say don't expect a raise or a bonus this year.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 48 points 7 months ago (5 children)

    In college, on the first day of orientation, someone in my class bragged that they wrote 50,000 lines of code for a game that was similar to tic tac toe, emphasizing that he "wrote a lot of code". A TA told him that it wasn't a sign that his program was decent and that it really didn't seem like it should take 50k lines of code to make something as simple as his game.

    He dropped out after the first week of intro to programming.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 26 points 7 months ago (4 children)

    That's nothing, I wrote the code to return if the input is even or not in 1M lines of code.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 17 points 7 months ago (3 children)

    If 1 no else if 2 yes else if 3 no....

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 17 points 7 months ago (9 children)

    I wouldn't say PR size is a bad metric, you usually just need yo read it the opposite of how sloppers do it, i.e. the most productive PRs are short and focused.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 9 child comments
    •
  • [–] 14 points 7 months ago (4 children)

    Then Devs focus on minifying the code into an unreadable mess

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 18 points 7 months ago*

    I'm not saying it's a good individual metric. In fact, applying individual metrics to developers (or most workers really), will only land you in Goodhart's hell.

    But as part of holistic operational health tracking, it's a useful team level metric, as there is ample evidence that shorter PRs tend to result in less operational issues. And, of course, this is only valid if you don't try to tie financial rewards to it, otherwise people will forget that PR size is a proxy measure for how easy changes are to review and rollback.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 7 months ago

    Yeah, I’ve been guilty of coming up with “cute” solutions that are extremely optimized and concise, but you needed to take a hundred times as long to work through what was going on.

    Usually I would put an explanation comment, but sometimes a less optimized solution is the better option for readability sake.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 7 months ago

    Hah, if those pesky devs think that they can play the system by just rolling up the code into a single line they got another thing coming - we're actually tracking PR character count, NOT LOC like some other companies!

  • permalink
  • fedilink
  • source
  • parent
  • [–] 5 points 7 months ago (2 children)

    PR size is an awful metric. The bigger the PR, the less reviewable it is.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 7 months ago

    I’m working with a legacy codebase for the last few months, where a simple PR often ends up crossing a 1000 lines count due to testing and commenting, and I can’t stop apologizing for those.

    Yet there are people out there bragging about 10x changesets.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 49 points 7 months ago (5 children)

    I'm just a hobbyist, but I'm always more proud of commits that remove stuff.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 32 points 7 months ago (3 children)

    Removing shit and it still working perfectly the same is absolutely a goal everyone should have. Less code means less to maintain.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 5 points 7 months ago

    You reminded me of a story I recently read, where the author highlighted just how much awesome programming someone had done by describing how their hands were cramping up.

    It's like estimating how well an artist paints by looking at how much paint is on their clothes, or judging how good a cook is by how many cuts and burns they have. The actions that cause those things are incidental to the process, not central, and an excessive amount points to incompetence, not hard and skillful work.

  • permalink
  • fedilink
  • source
  • parent