Topic: metal

3x faster speeds on MLX | Qwen 3.8 27B | Native MTP Speculative Decoding On Apple Silicon With No External Drafter.

Bend 2: a fast language that blocks AI mistakes via proof. Install: curl -fsSL https://bend-lang.com/install.sh | sh

3D engine with modern graphics

Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook

A local inference engine for Apple silicon, built around the model.

Run a 105 GB AI model on a Mac that can't hold it. Slotstream streams Qwen3.8-Flash-Next (125B mixture of experts) from your SSD and caches the busiest experts in memory, so it runs on Macs with 16 to 64 GB. One native Swift binary on MLX and Metal, no Python, offline. Works with Claude Code, Codex and Ollama or OpenAI clients.

Ocarina of Time native on iOS and iPadOS via Ship of Harkinian with Metal rendering, touch controls, controller support, and ROM-free reproducible builds.

41 hand-curated drop-in SwiftUI Metal shader effects — holographic foil, kaleidoscope, magnetic fields, glitch, neon edges and more. Typed View modifiers, agent-friendly docs.

A serious terminal emulator with a stupid name

The native macOS terminal that keeps your sessions running and tells you when a coding agent needs you. GPU-rendered, scriptable, agent-aware.

Bend 2: a fast language that blocks AI mistakes via proof. Install: curl -fsSL https://bend-lang.com/install.sh | sh

3D engine with modern graphics

Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook

3x faster speeds on MLX | Qwen 3.8 27B | Native MTP Speculative Decoding On Apple Silicon With No External Drafter.

A local inference engine for Apple silicon, built around the model.

Run a 105 GB AI model on a Mac that can't hold it. Slotstream streams Qwen3.8-Flash-Next (125B mixture of experts) from your SSD and caches the busiest experts in memory, so it runs on Macs with 16 to 64 GB. One native Swift binary on MLX and Metal, no Python, offline. Works with Claude Code, Codex and Ollama or OpenAI clients.