4 ms·

So you were generating and evaluating the performance of your CLAUDE.md files? And you got banned for it?

alistairSH 9mo ago
It reads like he had a circular prompt process running, where multiple instances of Claude were solving problems, feeding results to each other, and possibly updating each other's control files?
- epolanski 9mo ago
  What would be bad in that? Writing the best possible specs for these agents seems the most productive goal they could achieve.
  - NitpickLawyer 9mo ago
    I think the idea is fine, but what might end up happening is that one agent gets unhinged and "asks" another agent to do more and more crazy stuff, and they get in a loop where everything gets flagged. Remember that "bots configured to add a book at +0.01$ on amazon, reached 1M$ for the book" a while ago. Kinda like that, but with prompts.
    - epolanski 9mo ago
      I still don't get it, get your models better for this far fetched case, don't ban users for a legitimate use case.
  - alistairSH 9mo ago
    Nothing necessarily or obviously bad about it, just trying to think through what went wrong.
- andrelaszlo 9mo ago
  Could anyone explain to me what the problem is with this? I thought I was fairly up to date on these things, but this was a surprise to me. I see the sibling comment getting downvoted but I promise I'm asking this in good faith, even if it might seem like a silly question (?) for some reason.
  - alistairSH 9mo ago
    From what I'm reading in other comments, the problem was Claude1 got increasingly "frustrated" with Claude2's inability to do whatever the human was asking, and started breaking it's own rules (using ALL CAPS). Sort of like MS's old chatbot that turned into a Nazi overnight, but this time with one agent simply getting tired of the other agent's lack of progress (for some definition of progress - I'm still not entirely sure what the author was feeding into Claude1 alongside errors from Claude2).
- Hackbraten 9mo ago
  They were trying to optimize a CLAUDE.md file which belonged to a project template. The outer Claude instance iterated on the file. To test the result, the human in the loop instantiated a new project from the template, launched an inner Claude instance along with the new project, assessed whether inner Claude worked as expected with the CLAUDE.md in the freshly generated project. They then gave the feedback back to outer Claude. So, no circular prompt feeding at all. Just a normal iterate-test-repeat loop that happened to involve two agents.
Aurornis 9mo ago
I think it's more likely that their account was disabled for other reasons, but they blamed the last thing they were doing before the account was closed.
- pocksuppet 9mo ago
  And why wouldn't you? It's the only information available to you.

It reads like he had a circular prompt process running, where multiple instances of Claude were solving problems, feeding results to each other, and possibly updating each other's control files?

epolanski 9mo ago
What would be bad in that? Writing the best possible specs for these agents seems the most productive goal they could achieve.
- NitpickLawyer 9mo ago
  I think the idea is fine, but what might end up happening is that one agent gets unhinged and "asks" another agent to do more and more crazy stuff, and they get in a loop where everything gets flagged. Remember that "bots configured to add a book at +0.01$ on amazon, reached 1M$ for the book" a while ago. Kinda like that, but with prompts.
  - epolanski 9mo ago
    I still don't get it, get your models better for this far fetched case, don't ban users for a legitimate use case.
- alistairSH 9mo ago
  Nothing necessarily or obviously bad about it, just trying to think through what went wrong.
andrelaszlo 9mo ago
Could anyone explain to me what the problem is with this? I thought I was fairly up to date on these things, but this was a surprise to me. I see the sibling comment getting downvoted but I promise I'm asking this in good faith, even if it might seem like a silly question (?) for some reason.
- alistairSH 9mo ago
  From what I'm reading in other comments, the problem was Claude1 got increasingly "frustrated" with Claude2's inability to do whatever the human was asking, and started breaking it's own rules (using ALL CAPS). Sort of like MS's old chatbot that turned into a Nazi overnight, but this time with one agent simply getting tired of the other agent's lack of progress (for some definition of progress - I'm still not entirely sure what the author was feeding into Claude1 alongside errors from Claude2).
Hackbraten 9mo ago
They were trying to optimize a CLAUDE.md file which belonged to a project template. The outer Claude instance iterated on the file. To test the result, the human in the loop instantiated a new project from the template, launched an inner Claude instance along with the new project, assessed whether inner Claude worked as expected with the CLAUDE.md in the freshly generated project. They then gave the feedback back to outer Claude. So, no circular prompt feeding at all. Just a normal iterate-test-repeat loop that happened to involve two agents.

Aurornis 9mo ago

I think it's more likely that their account was disabled for other reasons, but they blamed the last thing they were doing before the account was closed.

pocksuppet 9mo ago
And why wouldn't you? It's the only information available to you.