Founder at RESON, Moebius, and Penrose NPO
Building emotionally steerable generative audio, local-first AI systems, and infrastructure for measurable creative workflows.
Generative audio · Real-time steering · Emotional control surfaces · Local-first AI systems · Reproducible infra
Current work · Open source · Technical focus · Principles · Selected builds · Stats
Public proof links for recent AI tooling and TUI ecosystem work:
| Project | Proof |
|---|---|
| ratatui/ratatui | PR #2555 · fork |
| openclaw/openclaw | PR #84995 · fork |
| openai/codex | Issue #23887 · fork |
| earendil-works/pi | Issue #4868 · fork |
Emotionally steerable generative audio for continuous, personalized sound flows.
- Real-time mood steering across Focus, Energy, Calm, and Joy
- Continuous audio generation instead of static playlist selection
- Browser audio reliability, DJ-style transitions, and exportable listening sessions
An organizational operating system focused on digital twins, decision intelligence, and project health metrics.
- Emphasis on structure, traceability, and long term scalability
- Designed for durable decision making, not short term automation hacks
- Built to expose organizational reality through models, metrics, and constraints
Languages
- Rust (primary)
- TypeScript
- Python
Domains
- Applied AI systems and orchestration
- Generative audio and realtime interaction
- Distributed backends and data models
- Product and organizational metrics
- Web applications with strong architectural constraints
Stack
- Rust backends
- Tauri, Vite, modern web tooling
- AI model integration (local and hosted)
- Clarity over convenience
- Traceability over vibes
- Constraints as a feature
- Systems that scale in time, not just in traffic
- Metrics that explain, not merely report
|
Emotionally steerable generative audio for realtime mood-shaped listening.
|
Organizational OS for digital twins, decision intelligence, and project health metrics.
|
RESON thesis
- The listener should shape a state, not browse a catalog.
- The interface should expose emotional intent as a controllable signal.
- Generated audio should remain continuous, steerable, and explainable through session data.
- The long-term model is contextual generation informed by psychoacoustics and feedback loops.
Moebius philosophy
- The goal is not to automate people.
- The goal is to make organizational state explicit, inspectable, and stable.
- Systems fail when decisions are untraceable and metrics are disconnected from structure.
- Moebius treats structure as a first class artifact, then builds AI around it.
What I mean by “digital twins”
A digital twin here is a structured representation of an organizational entity.
- People, roles, teams
- Processes, projects, constraints
- Decisions, rationales, measurable health indicators
The twin is only useful if it is:
- grounded in real artifacts
- consistent over time
- able to explain why the system recommends something
Tooling interests
- Developer experience that reduces ambiguity
- Interfaces that enforce clear input and clear output
- Local first ergonomics where it matters
- Evaluation and measurement pipelines for AI behavior