Independent software lab building interactive reading experiences and inspectable agent systems.
We make an instrument for reading, a workspace for thinking with AI over live data, and infrastructure that lets agents do work under human keys.
RISE is a browser-based audiovisual reader. It presents text through time, image, sound and procedural visuals, so a book can be read as a timed stream or a typeset page. Open beta, Apache 2.0.
Open RISE → · About the project · Architecture
| Project | What it is |
|---|---|
| OmniOS | A canvas for thinking with AI over live data. Blocks pull real numbers, wires feed them into personas, and a question is answered from what the wires carry. It runs without keys on public sources and local models. A capability engine compiles OpenAPI and MCP descriptions into typed blocks; writes wait for approval. |
| SyberWork | Contracts govern work. A case history records what was proposed, admitted, observed, approved and executed, in a hash-chained, signed event log. Apache 2.0. |
| OSAHR | A Python kernel for exact, open, stochastic adaptive rewriting of typed directed hypergraphs, with event replay and no runtime dependencies. Apache 2.0. |
| Papers | Research artifacts at different stages, each with its evidence, claim limits and reproduction steps. |
Engines is Mateo's generative work: seven procedural engines, each a single HTML file in Canvas 2D or WebGL2, no libraries.
Agents write much of our code. People hold the keys: nothing an agent produces is admitted until a person, or a gate a person set, accepts it. Every repository states what is implemented, what is partial, and what its results do and do not show.
- Mateo Robles — product, HCI, machine learning and systems
- Seth Carlson — engineering infrastructure and architecture on RISE; systems work on OSAHR
