2 pointsby ABS2 hours ago1 comment
  • ABS2 hours ago
    among the many things I found that were not highlighted anywhere (that I could find) is the distributed inference across Macs (JACCL over Thunderbolt 5), an OpenAI-compatible mlx_lm.server, agentic-on-Mac.

    Apple keeps MLX (bring-your-own-weights) separate from Foundation Models / Core AI.