Pinned Loading

1. ane-infer ane-infer Public

   Apple Neural Engine (ANE) LLM inference engine — reverse-engineered private APIs, Metal GPU shaders, hybrid ANE+GPU+CPU on Apple Silicon. 32 tok/s matching llama.cpp, 3.6 TFLOPS fused ANE mega-kern…

   Rust 29 5

2. brainbox brainbox Public

   Muscle memory for Claude, OpenClaw, and AI agents. Zero-cost Hebbian memory system — learns which files matter through co-access patterns, predicts what you need next.

   TypeScript 13 3

3. ane-synth ane-synth Public

   Real-time neural synthesizer on Apple Neural Engine — bypassing CoreML, 157μs latency, 79x real-time

   Rust 10

4. ssd-apple-silicon ssd-apple-silicon Public

   SSD (Speculative Speculative Decoding) on Apple Silicon M4: comprehensive research showing why 66.9 tok/s MLX spec decode is the ceiling

   Swift 3 2