645
Classic (lemmy.ml)
submitted 10 months ago by to c/programmerhumor@lemmy.ml
59 comments fedilink hide all child comments
 
you are viewing a single comment's thread
view the rest of the comments
[–] 113 points 10 months ago (12 children)

Missing the LLM developer with billions of tokens to burn.

  • "Claude, please solve this puzzle so my Grandma can get her insulin"
  • if the LLM succeeds, solves in less than 10 seconds.
  • if it fails, star remains forever unsolved - "Impossible with current LLM, will try again with ChatGPT 8.0"
  • permalink
  • fedilink
  • source
  • hideshow 12 child comments
    •
  • [–] 15 points 10 months ago (7 children)

    Missing the LLM developer with billions of tokens to burn.

    developer

    developer

    are they tho?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 7 child comments
    •
  • [–] 10 points 10 months ago (6 children)

    script-kiddie 2.0?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 3 points 10 months ago (5 children)

    Close. A script is readable and editable code. The LLM slop is "deep" so it's opaque and you can't really change the system prompt (you can jailbreak it).

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 5 points 10 months ago (4 children)

    Script-kiddie is a derogatory term for a computer user who can't read or edit the code. Scripts might be useful to us, but they may as well be opaque to a script-kiddie.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 1 point 10 months ago (3 children)

    Would you really claim that a script-kiddie won't alter a number or a string in some script?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 1 point 10 months ago* (2 children)

    For my definition of script-kiddie, its not that they won't change something, they by definition cannot because they dont understand it.

    We may have different definitions here, but that is the one I am familiar with.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 10 months ago* (1 child)

    I'd say a script-kiddie can modify a script, but only if there's some resource that says 'change this part to do x'. They don't understand why that change does what it does, though, or how to troubleshoot if it doesn't actually do the thing.

    Signed: config file kiddie - I just do what youtube or forum denizens tell me to do until it works. Imagine there's script-kiddies out there that do something similar.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 10 months ago

    there’s some resource that says ‘change this part to do x’.

    Arguably, this is just a different type of script.

    I think the real separation is that while you might seek out a few different resources, or try a few alternatives to solve a problem, a "script-kiddie" is likely stuck if the first result in Google doesn't solve the problem immediately.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 6 points 10 months ago
    • Spends the first 90% of the competition developing specialized subagents and custom MCP servers to allocate the problems and most relevant information efficiently into the LLM's contexts.
    • All of his agents easily escape their own sandboxes and one accidentally configures itself into "delete-only mode".
    • "Codex, how the fuck do you not have access to your own documentation?"
    • Places 29th globally after one of his subsubagents finds a way to reconstruct the full solution set from filesystem metadata in the online judge VMs.
  • permalink
  • fedilink
  • source
  • parent
    • load more comments (1 reply)