I watched Alex Karp’s recent talk at Davos/WEF where he stated: "Half your enterprise doesn't work on the battlefield. It works in a lab on a PowerPoint."
As a self-taught architect fighting German bureaucracy, I felt that. So I built an engine to fix the "Execution Gap" between LLM capabilities and strict operational constraints.
The Solution: CausaNova I built a deterministic .NET 8 runtime (Zero-Dependency) that acts as a firewall between the "Intent" (AI) and the "Action" (Reality).
It is already domain-agnostic. While the demo shows a kinetic defense scenario, the engine doesn't care about the target. We have already implemented ontologies for:
HTML: Rendering strict UI components.
SQL: Enforcing safe query states.
Math: Solving geometric constraints.
PLC/Hardware: controlling signals. To CausaNova, these are just domains in a JSON DSL.
How it works:
Logic over Code: We stripped business logic out of the C# code entirely.
Ontology as Config: Rules are defined in a strict JSON schema.
The Resolver: The runtime uses an internal solver (SMT-style logic) to validate every request against the current state before execution.
Tactical Edge: No cloud required. Runs on low SWaP hardware.
Why? Because "Ethics" is a debate, but "Logic" is a binary state. I wanted a system where I can hot-reload the "Rules of Engagement" (JSON) without recompiling the kernel.
The Stack: .NET 8 (AOT), Zero external dependencies.
I’m looking for feedback on the ontology structure and offline-inference strategies. Roast my architecture.
Hello HN,
I watched Alex Karp’s recent talk at Davos/WEF where he stated: "Half your enterprise doesn't work on the battlefield. It works in a lab on a PowerPoint."
As a self-taught architect fighting German bureaucracy, I felt that. So I built an engine to fix the "Execution Gap" between LLM capabilities and strict operational constraints.
The Solution: CausaNova I built a deterministic .NET 8 runtime (Zero-Dependency) that acts as a firewall between the "Intent" (AI) and the "Action" (Reality).
It is already domain-agnostic. While the demo shows a kinetic defense scenario, the engine doesn't care about the target. We have already implemented ontologies for:
How it works: Why? Because "Ethics" is a debate, but "Logic" is a binary state. I wanted a system where I can hot-reload the "Rules of Engagement" (JSON) without recompiling the kernel.The Stack: .NET 8 (AOT), Zero external dependencies.
I’m looking for feedback on the ontology structure and offline-inference strategies. Roast my architecture.