ashhart/TensorFold
Fast, exact LLM decoding on Apple Silicon (MLX) behind an OpenAI-compatible endpoint
Independent AI systems architect building local-first AI tools and governed inference infrastructure.
Repositories
ashhart/MCDMA
Metal Cuda Direct Memory Access
ashhart/homebrew-tensorfold
Homebrew tap for TensorFold: brew install ashhart/tensorfold/tensorfold
ashhart/SparkPilot
SparkPilot puts your NVIDIA DGX Spark to work, free forever.
ashhart/Imprint
Speeds up TTFT
ashhart/OpenBot
OpenBot - A reimplementation of GrokBot
ashhart/DGX-Anti-OOM
A small script to stop your DGX Spark OOMing
ashhart/ComputerUse
Use computer use with tools like OMP or OpenCode.
ashhart/Drift
We gave AI models telepathy.
ashhart/Syntra
Decision engine
ashhart/Detectra
Local AI Chrome Plugin for AI Image Detection.
ashhart/mlx
MLX: An array framework for Apple silicon
ashhart/Cortheon
A lightweight runtime that helps a small local model reason, discover, and complete work like a frontier model.
ashhart/codex-local
Use Codex with local models.
ashhart/Duo
Two models, one workspace: an OMP plugin for shared tasks and Agent Hub conversations.
ashhart/omlx
LLM inference server with continuous batching & SSD caching for Apple Silicon — managed from the macOS menu bar
ashhart/Qwen38-Flash-Next-MLX
Qwen 3.8 Flash Next 4bit MTP running at just under 100tks decode
ashhart/silicone-exchange
Benchmarking Alex Ziskind's Silicone Exchange with DGX Sparks
ashhart/ashhart
AshHart
ashhart/mlx-swift-lm
LLMs and VLMs with MLX Swift
ashhart/mlx-swift
Swift API for MLX
ashhart/d-inference
Private Inference Network on Idle Macs
ashhart/DeepSeekV4-Flash-0731-MXFP4-MLX
DeepSeekV4-Flash-0731-MXFP4-MLX Apple Silicone