Notes on working with language models.
These pages are about the ordinary choices that show up once a model is part of the work: where a context budget actually goes, what a small test set hides, and why a prompt is a poor place to keep a boundary.
These pages are about the ordinary choices that show up once a model is part of the work: where a context budget actually goes, what a small test set hides, and why a prompt is a poor place to keep a boundary.