R.E.N.
Reasoning Engine Network
An orchestration engine we're building in-house, designed so that the system, not the AI model, is the authority on what happened.
Early stage
Research & Studio
Early ideas, tested before they become something larger.
The Studio is where Nasar Labs explores ideas before deciding what they become.
Some experiments become technology. Some become ventures. Others simply teach us something and stay on the bench.
Our work spans intelligent systems, software architecture, engineering and emerging technologies. We are interested in the difficult parts: how systems reason, how they interact with the real world, and how ideas can be turned into dependable technology.
Reliable agentic systems
When the model cannot be the source of truth.
We are exploring how intelligent software can reason and propose actions without allowing the language model itself to become the authority on what actually happened.
The model may propose. The system must assert.
Authoritative state should come from verified system data, deterministic operations and explicit permissions, not from what a model believes happened.
This work informs how we approach dependable AI systems at Nasar Labs.
Ongoing research
Systems and infrastructure
Building the foundations before building on top of them.
Our experiments often begin beneath the visible product: databases, system boundaries, authentication, infrastructure, orchestration and the interfaces between intelligent software and the systems it operates.
We treat these foundations as part of the product itself.
Ongoing
The Studio
Not everything we build belongs here.
We publish work when there is an idea worth sharing, an engineering lesson worth documenting, or a system mature enough to stand on its own.
The rest stays on the bench until it is ready.
More experiments and technical notes will be added as the work develops.