Building private, local-first AI products for voice and developer workflows.
Projects · Open source · Principles
I make AI tools useful in the real world: local data paths where possible, explicit privacy boundaries, reproducible engineering work, and interfaces people can trust every day.
Auris — private meeting memory for Windows
Local-first call recording, on-device Whisper transcription, and AI notes through an OpenAI-compatible provider chosen by the user. Auris records microphone and system audio on separate tracks, then turns conversations into useful summaries and Q&A.
Rust · Tauri · React · TypeScript · Whisper · SQLite
Explore Auris · Download a release
Agent Workflows — reliable prompts for coding agents
Portable Markdown workflows for three high-value moments: preserving session context, auditing a project safely, and interviewing an idea into a buildable plan. Includes an installable Codex plugin and tool-neutral prompt files.
Browse workflows · Install the plugin
bugkit — context that helps AI fix bugs
An AI-friendly bug-report workflow: users capture rich context, maintainers receive one concise technical brief, and coding agents can start investigating without losing the important details.
CipherBoard — offline-first secure keyboard for GrapheneOS
An encrypted Android keyboard built around physical QR pairing, Olm Double Ratchet, Android Keystore, and no network permission.
Explore CipherBoard · View releases
- browser-use #5217 — preserve accessibility-name semantics in enhanced DOM handling.
- prompts.chat #1223 — add validation and sensible bounds to public API pagination.
I prefer a small number of scoped, reproducible contributions with tests and clear maintenance value over activity for its own sake.
- Local-first by default. Keep sensitive audio, transcripts, and context with the user whenever possible.
- AI should be composable. Let people bring an OpenAI-compatible endpoint or key instead of locking them into one provider.
- Evidence beats theatre. Ship narrow changes, show validation, document trade-offs, and leave a project easier to understand than before.
Making Auris a dependable private workflow for recording, transcription, and useful AI notes; expanding Agent Workflows with reusable, safe patterns for coding agents.
Good feedback, focused issues, and thoughtful collaboration are always welcome.