Discussion
Loading...

Post

Log in
  • About
  • Code of conduct
  • Privacy
  • Users
  • Instances
  • About Bonfire
Zef Hemel
Zef Hemel
@zef@hachyderm.io  ·  activity timestamp 2 days ago

I think issuing 20+ #AI generated PRs in a timespan of 24h can be considered the open source equivalent of a #DoS attack right? Sounds like a reason to ban without explanation?

https://github.com/silverbulletmd/silverbullet/pulls

That would be the sensible thing to do, but I’m not sensible so I’m going to overcook and probably spend hours this weekend writing an article in which I explain the hostility of this use of AI. Because I want to fix the world.

  • Copy link
  • Flag this post
  • Block
Brian David
Brian David
@ectopod@hachyderm.io replied  ·  activity timestamp yesterday

@zef Are these PRs related to an existing issue, or are these more "finding" an issue and then "fixing" it? Not that it makes it any less frustrating, but the latter feels significantly more harmful.

  • Copy link
  • Flag this comment
  • Block
Zef Hemel
Zef Hemel
@zef@hachyderm.io replied  ·  activity timestamp yesterday

@ectopod some are supposedly fixing real issues, yes: https://github.com/silverbulletmd/silverbullet/pull/1750
https://github.com/silverbulletmd/silverbullet/pull/1747

Always obscure ones, though. And I think a bunch more come from what I suspect are prompts like "fix some common security issues in this project"

  • Copy link
  • Flag this comment
  • Block
Zef Hemel
Zef Hemel
@zef@hachyderm.io replied  ·  activity timestamp yesterday

Ok, cooled down a bit. Closed all PRs and left a reasonable note (IMO): https://github.com/silverbulletmd/silverbullet/pull/1750#issuecomment-3694069185

  • Copy link
  • Flag this comment
  • Block
Mark IJbema
Mark IJbema
@mark@tacobelllabs.net replied  ·  activity timestamp 23 hours ago

@zef love how a well written "no" makes the discussion more productive (at least his response sounds quite reasonable)

  • Copy link
  • Flag this comment
  • Block
Bas (Tools on Tech)
Bas (Tools on Tech)
@toolsontech@pkm.social replied  ·  activity timestamp yesterday

@zef words well chosen, AI patches really don't take the real world into account where people will have to deal with the resulting issues.

  • Copy link
  • Flag this comment
  • Block
Zef Hemel
Zef Hemel
@zef@hachyderm.io replied  ·  activity timestamp 19 hours ago

@toolsontech this is one of many unintuitive reasons why AI may actually slow software development down.

  • Copy link
  • Flag this comment
  • Block
Bas (Tools on Tech)
Bas (Tools on Tech)
@toolsontech@pkm.social replied  ·  activity timestamp 9 hours ago

@zef From what I have seen the benefits are mostly when a capable person uses it to do busywork. The drawbacks usually happen when supervision is gone and it's all automated.

And I'm not even starting on the security risks that are currently vibe coded

  • Copy link
  • Flag this comment
  • Block
Marten
Marten
@marten@mstdn.social replied  ·  activity timestamp yesterday

@zef also really obnoxious to include those server tests in everything, even if they don't matter for the PR at hand. Eg 1749 doesn't include claude as co-author for the principal commit and seems reasonable to my (non-go programmer) eyes but i would never merge it because of the 950 lines of unrelated server tests

  • Copy link
  • Flag this comment
  • Block
A
A
@a@852260996.91268476.xyz replied  ·  activity timestamp 2 days ago

@zef@hachyderm.io do you know what is the motivation of the issuer? I'm honestly curious.

  • Copy link
  • Flag this comment
  • Block

bonfire.cafe

A space for Bonfire maintainers and contributors to communicate

bonfire.cafe: About · Code of conduct · Privacy · Users · Instances
Bonfire social · 1.0.1-alpha.40 no JS en
Automatic federation enabled
Log in
  • Explore
  • About
  • Members
  • Code of Conduct