Building AI tools and systems that hold up in daily use!
Building now: cortex · gpr · Model-Switchboard
If you are exploring agent tooling or inference, these are the main repos.
Persistent shared memory and context compression for AI coding agents. Rust daemon with HTTP, MCP, and a desktop control center.
Rust · MCP · agent memory
Model routing for Mac. Pick models and switch quickly across local and hosted backends.
Python · model routing · Mac
Goal-driven PRD ratchet for agent workflows. Work stays open until real artifacts pass checks.
Python · spec-driven · agent workflows
| Project | What it is |
|---|---|
| RustLibraries | Python libraries ported to Rust |
| TweetKB | Organize and analyze Twitter bookmarks |
| TwitterArticles | Chrome extension to download Twitter articles |
Inference workbenches (forks and experiments)
| Project | Notes |
|---|---|
| llama.cpp | C/C++ LLM inference workbench |
| lucebox-hub | Hand-tuned inference for consumer hardware |
| rvllm | Rust LLM inference / vLLM-style serving |
| vllm | High-throughput LLM serving |
- Use local, cloud, or hybrid setups depending on what the work needs.
- Prefer artifacts, tests, and clear completion checks over vibes.
- Move fast on experiments; move carefully on published work.
- Treat agent systems as infrastructure, not demos.
