Re: Mailing list AI checking tool

[Rob Sayre <[e-mail]>]

Hi,

Not sure if it's on purpose, but this is exactly what Claude would write:

"Honest caveats"

https://github.com/JayDaley/mailing-list-ai-check

The AI stuff is ok, but I've always objected to mixing human and machine
traffic, because the machines don't understand the cost for the humans.

I treat the AI traffic like reading logs or something.

This is in my CLAUDE.md:

### Interactions

Do not suggest "giving up for the day". The authors will just give up on
their own.

Drop "honest"/"honestly" as a framing word — no "the honest upshot", "to be
honest", "honest synthesis", etc. Honesty is assumed; labeling something as
honest just adds noise and implies the rest might not be. State the point
directly.

###

It's really in there from a few months ago, on a giant graphics project
unrelated to IETF work.

thanks,
Rob