Turn your Mac into the brain of your household and operate from your phone.
Educational distributed training and inference library for heterogeneous local hardware: Mac minis, Raspberry Pis and GPUs over plain Python sockets. FSDP, GRPO, parameter-server, expert and data parallelism.
Self-hosted AI platform on Apple Silicon — production-proven, Thai-optimized, $3.50/day
Open a browser tab — you're now driving your computer. Live remote desktop + real PTY shell (Claude Code, vim, top) + on-device voice. macOS and Windows, every machine you own behind one URL and one password. Self-hosted.
Prajna — on-device memory-augmented reasoning: a 6.7M Cognitive Resonance Network inside frozen Gemma 4 E2B. Passes CEHRI 60/60 via retrieval memory, 91.7% on unseen reworded exams. Trained on a Mac Mini M4, no GPU, no API.
GitHub Actions cron that watches Apple's refurb store for an M4 Mac mini under a price cap and pings Slack.
Notes from running AI on my own Mac mini: model speeds, what broke, the benchmark script, and people building local AI.
NotchX is adds MacBook like notch to Windows ,Linux and Mac Desktop devices at the same time providing better user experience
Memory-safe, LAN-accessible OpenAI-compatible server for Gemma 4 12B (4-bit MLX) on a 16 GB M2 Pro Mac mini
Decode speed of 16 local LLMs on a Mac mini M6 (32 GB) in LM Studio: llama.cpp vs MLX vs Splash speculative decoding, October 2026. Raw data and script.
Blind, task-oriented eval: can a local LLM actually WRITE as well as a frontier model? Measures output quality, not tokens/sec.
A platform-agnostic format + method for sizing local-LLM hardware from your real agent sessions, calibrated with capability-oracle runs.
Give your local AI eyes and memory — native to your Mac. Apple Silicon LLM (MLX) + web proxy + conversation RAG in one process, OpenAI-compatible.
Free, self-hosted, ultra-low-latency Push-to-Talk voice typing & AI translation for Windows, offloaded to an Apple Silicon Mac via Apple MLX & Ollama.
MacBook + Mac mini as one system: your laptop is the console, the mini is the always-on runtime. SSH wiring, cron that actually fires, and a watchdog that makes silence trustworthy.
Headless Mac mini fan control bridge: spend cheap fan life to protect expensive storage thermal conditions.
Opens an AI agent session in Herdr for each Soma task marked for dispatch
🤖 Personal AI stack on Mac: Ollama + Open WebUI + Stable Diffusion + TripoSR 3D + Whisper STT + Tax AI Social + Gmail AI + Quant AI — 100% local, no cloud
Local-first assistant for safely capturing, classifying, and organizing Telegram/OpenClaw inputs in an Obsidian vault.
To associate your repository with the mac-mini topic, visit your repo's landing page and select "manage topics."