100k … I hear and use all of these high token context LLMs, but often they fail to include all information that should be in their context.

Does your approach work for you?