#local-ai
Wiki 2
- Mesh LLM: Distributed Inference over iroh Pools GPUs across machines as one OpenAI-compatible API; splits big models across nodes over iroh's p2p QUIC transport
- Preferring Local OSS LLMs Neward's case for local open-source LLMs: cloud fragility, economics, privacy, and learning
Toolbox 4
- CutItOut Browser background remover that runs U2-Net in WebAssembly, so images never leave the machine
- Ghost Pepper Hold-to-talk macOS dictation using on-device WhisperKit plus a local LLM cleanup pass
- Kokoro 82M-parameter text-to-speech model that runs fast on CPU, with ~50 voices across three languages
- meetily Meeting assistant that records, transcribes with Whisper or Parakeet, and summarizes locally