...and I usually come to doubt my own intuitions that this is the case when people say things like this, but my experience is usually that the LLM is doing more heavy lifting than you realise.

> Distill - deterministic context deduplication for LLMs. No LLM calls, no embeddings, no probabilistic heuristics. Pure algorithms that clean your context in ~12ms.

I simply do not believe that this is human-generated framing. Maybe you think it said something similar before. But I don't believe that is the case. I am left trying to work out what you meant through the words of something that is trying to interpret your meaning for you.