Context engineering: shifting from "tokenmaxxing" to deliberate curation

I've also build a solution that uses llm council, loops, context engineering, etc, fully open source, if you're interested - https://github.com/looptroop-ai/LoopTroop