Coding agent with a memory of all things

Open any development server on any device in your tailnet

Lossless greedy speculative decoding for 4-bit Qwen3.8-27B on one RTX 3090, with the acceptance logic proven in Bend

The fastest decode and prefill for 4-bit Qwen3.8-27B on one RTX 3090 that preserves the model's intelligence

Turn one SVG into optimized SVG, PDF, favicon.ico, and PNG, WebP and JPEG from 16 to 4096 px