Comment by mediaman
1 month ago
I have seen this happening with contracts. Using AI to help evaluate contracts is fine, but I'm getting 5-10 page Claude docs with dozens of asks that they haven't read and many of which don't make sense. I find it pretty counterproductive, because it makes negotiation almost impossible -- you can't tell what the other side wants, because they haven't even read the output.
Or, they never actually sat down to THINK about it, which is even worse.
Yes, that's the root of it. Just let Claude create a list of demands. Then you've "done the job" of contract negotiation. You don't even have to think about what the business needs or what the partnership goals are.
It matches the pattern of LLMs being very good at simulating the form of work output, which is an issue with code but it seems quite exacerbated with anything non-verifiable, like written communication.
This