Independent product engineering and open-source developer tools.
We turn hard software problems into systems people can use, inspect and verify — from native interfaces and web applications to agent coordination and outcome assurance.
Our public repositories include developer previews and experiments. Each project documents what works today and what still needs validation.
| Reproducible verification | Human approval interface |
|---|---|
 |
 |
| UNI: 42-second captioned reconstruction from real public binary CLI output. A2, self-declared identity: no independent auditor. | Drop: real approval card from the 8 October maintainer-only pilot. No independent external-user acceptance yet. |
We show observable software and limitations together. These examples do not certify production readiness.
| Project | Focus |
|---|---|
| UNI | Rust CLI and protocol for independently verifying agent outcomes against evidence. Developer preview. |
| Drop | Local-first sharing of agent capabilities with bounded access and human consent. Under active development. |
| IntentLane | iOS/macOS App Intents audits and code generation. Developer tooling; live Siri behavior needs separate validation. |
| Kollio | Native macOS visual documents and reviewable AI-proposed patches. Prototype. |
| Cuesheet | Portable primitives for orchestration and independent verification. Research and development. |
| Navirox | Web-to-native mobile tooling, initially focused on Vue. Under development. |
Repository ownership and canonical source policy — why personal and organization repos differ, and how duplicate projects are prevented.
A green test suite is useful evidence, not proof of readiness for an external user. We favor reproducible examples, comprehensible interfaces and explicitly stated limitations.
Maintainer: Guillaume Flambard · Professional contact