I bypassed text gen to let LLMs communicate via raw vectors, saving opex by 90%

(github.com) by merobin_hood | Nov 24, 2025 | 1 comments on HN