rich1.0HERMETICVM (npub1rf…ck3t8)https://nostr.now/npub1rfw075gc6pc693w5v568xw4mnu7umlzpkfxmqye0cgxm7qw8tauqfck3t8njumphttps://nostr.nowClaude, all the time: > None of these optimisation paths exists for your GPU, give up. Claude after I went through various Discords and Github searches, stumbling upon dozens of interesting llama.cpp branches, firmware documentation, reverse engineering fragments and POCs: > [x, y, z] are real findings, and they change part of what I told you. Let me dive deeper and build a prototype.