
Building Omnitome: How I Squeezed a Multi-Process AI Reading Engine into 4GB VRAM
Loss-bounded semantic compression, 100% deterministic source citations, and conquering 14.6GB RAM explosions on a consumer laptop.
Deep-dive engineering notes on local-first AI, edge neural inference, retrieval architectures (RAG), and memory-constrained desktop runtimes. Building and shipping open-source software that runs privately without cloud egress.