Show HN: MESI cache coherence for multi-agent LLM context sync – 95% less tokensgithub.com/hipvlady1 pointhipvlady2 months ago