> Metadata is open-world positive evidence; missing facts do not prove SQL invalid. Numeric exactness, JSON/temporal profiles, and nested containers remain separate evidence boundaries.
right
right
Feel like latest models have really got hung up on “facts” and “evidence” in a really unnatural way.
And especially Claude seems to have a lot of trouble separating it’s internal thoughts, from prompts/instructions, to output.
So you get all of these weird comments and artifacts in the output.
“neighboring versions do not inherit certification. No tag, npm publication, Pages deployment, or release authorization is implied.”