Document how to create a benchmark for AI alignment

[emmairwin]

How can we take rubrics, metrics, experience and create benchmark - documentation would be democratizing to help people see anyone can contribute, and build benchmarks in their community/tech/area of knowledge

Use this issue to dump knowledge, ideas, examples

[emmairwin]

benchmarks can be positive (do this, or negative) don't do this.

• and then to be clear why is it bad, why is it bad

[MoralCode]

Benchmark like, an example of a moderation standard projects can use as a starting point? Or more like measuring/scoring AI contributions by quality/having a series of checkboxes looking for anti-patterns (like darts package repository does with https://pub.dev/packages/csafe_fitness/score (scoring/points for helpful actions taken)