Orbital

Origin

Real-world-asset tokenization deal structuring is partner-time-expensive, slow, and dependent on a small number of human attorneys and bankers. Each deal is a one-shot artifact; nothing about the structuring process compounds across deals. Adjacent markets — carbon, biodiversity — failed for the same reason: no oracle was held accountable for any prediction made before any outcome resolved. Orbital was built as the engine that produces verifiable receipts at every transition of every deal across every jurisdiction in scope.

Problem

Replacing one-off deal structuring with engine-level structuring requires three things at once: a deterministic compliance layer that routes the cheapest legal path through a graph of jurisdictions and asset classes; a probabilistic strategy layer that judges which structuring choice is right when more than one is technically valid; a receipt layer that anchors every transition cryptographically so the engine compounds intelligence from its own deal flow. Most tokenization shops solve one of these and pretend the other two will follow.

Approach

The Separation Principle. Computers verify structure — deterministic, instant, binary: explicit compliance rules, hash-chained receipts, weighted-graph pathfinding across jurisdictions, evidence-gated state-machine transitions. AI judges strategy — probabilistic, cheap per call: a multi-agent deliberation pipeline with adversarial review on top. The two run side by side; neither tries to do the other's job. Layered on top: an evidence base, a compliance lattice, a deal state-machine, a lawpack engine, the living instruments themselves, and a settlement corridor.

Methodology

Relational throughout. The system inventory is itself queryable: every jurisdiction, every corridor, every regulatory domain, every asset-class weight modifier, every evidence entry, every deal, every receipt sits in structured rows with explicit relationships. The deal-document generation pipeline runs continuously, scored on quality, audited on outcome. Every deal can be tested, audited, and inherited; the engine has demonstrated, end-to-end, autonomous analysis of its own state followed by code-level proposals to itself with no human in the loop — the human is at the approval boundary, not the typing boundary.

Selected milestones

• Engine layer operational; receipt chains, compliance routing, and deal state-machine live in production
• A-grade pipeline spans treasuries, real estate, funds, bonds, on-chain yield, private equity, and private credit
• Self-improvement loop demonstrated: autonomous analysis to code-level proposals, zero human intervention in the production step
• Multiple sovereign jurisdictions and asset classes wired into the routing graph

Open questions

• How the tensor's regulatory dimensions should extend as new asset classes enter
• Where to anchor receipt chains for highest-leverage settlement
• Whether the deal-routing graph should converge with the standalone jurisdictional compliance graph or stay deliberately scoped
• How the engine should weight new evidence when an asset class has both A-grade and D-grade exemplars

Ask me about

• What the Separation Principle means for compliance work in practice
• How the tensor self-improves from deal outcomes
• Why yield-stripping methods are systematically underused in on-chain tokenization
• What separates an opportunity score from a document-quality score in deal generation
• How autonomous code-proposal works — and where the human-approval boundary sits

Stack