This is your work, valued
web-to-app. The most full featured web-to-app toolkit on Android, a complete APK workshop that runs entirely on your phone
5.4kWebToApp. Turn any website into an installable app across iOS, Android, Windows, macOS and Linux.
37jseek. The fastest way to pull values out of JSON in Go — zero allocations, no structs, no full parse
30turn-mcp. Achieve infinite conversation turns in a single API request via turn-mcp. Self-hosted MCP server with browser console for human-in-the-loop AI agents.
26webtoapp-releases. webtoapp-releases
6wgsql. GPU-accelerated columnar OLAP via WebGPU. One WGSL kernel, four backends.
6zv-codec. Tiny video/audio codec optimized for screen recordings. 24% smaller than x265 lossless RGB at the same speed, ~1100 lines of C, no NN, no external deps.
5MacBottle. Swift
4echomind. A non-attention LM recipe with proven small-scale ICL and Chinchilla-style scaling law (R²=0.995). Reproducible in 25 minutes on a MacBook.
4anthropic-ui-skill. A Skill that allows AI to generate the UI design style of Anthropic Company. Its aesthetics are very good.
3qwen2.5-1.5b-oasst-recipe. Example mlrecipe: Qwen 2.5 1.5B + 7-module rank-16 LoRA. 74MB recipe rebuilds a ~3GB merged checkpoint bit-identical to PEFT.
3mlrecipe. Content-addressed format for LoRA fine-tunes. Recipe + hashed adapter rebuilds the merged checkpoint bit-identical to PEFT.
2gpt2-alpaca-recipe. Example mlrecipe: GPT-2 + Alpaca LoRA. 1.1MB recipe rebuilds a 500MB merged checkpoint bit-identical to PEFT (148/148 tensors, max diff 0).
2revx. Rust reverse-engineering toolkit with CLI and local MCP server.
1memx. MemX — ~100× lower LLM memory on Apple, original precision. Stable planes: ~9× phys · ~100× engine · ~300× vessel capability (bitexact).
1