Context Loom is a small independent practice focused on one thing: the material you put in front of a language model, and how to get it right.
We are not a model shop and we do not sell a platform. Most of the work is unglamorous. We look at what a system is actually feeding its model, we find the duplicates and the contradictions and the stale pages, and we help teams build a pipeline that decides what earns a place in the window.
These notes are written as we go. Some are short observations from a single afternoon of debugging. Others are longer guides that we point clients to so we do not have to explain the same idea twice. If something here saves you a bad week, that is the intent.
You can read the field notes for the short pieces, or the guides for the longer ones.