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

Story dated 2026-09-30 by the Algolia API; 194 points and 97 comments at access 2026-10-04. Useful as an adversarial check on engine consensus: several commenters independently nominate MLX rather than llama.cpp as the correct Apple baseline, and one reports being 20-30% faster on llama.cpp than the launch product on the same NVIDIA box. Points and comment counts are a popularity proxy, not a measurement.