Launch HN: Magnitude (YC S25) – Self-optimizing inference engine for agents

Magnitude is an open-source, self-optimizing inference engine for agents that runs on Mac, Linux, and Windows, and is up to 2x faster than llama.cpp. It uses on-device compilation and tuning, focuses on best architectures, and has dynamic memory allocation and hybrid paged attention. It is built in Rust and is designed to run local agents with long sessions, allowing users to use their computer for other tasks. Magnitude is fully open-source and is available as a desktop app that can connect with various agents. The developers plan to improve the engine with expert streaming, a kernel compiler, and multi-device utilization.