Nakatomi Systems
Nakatomi Systems
Each entry below is a filed method. We describe what it does, who it serves, why it matters now, and what early engagement secures. The mechanisms live in the filings, and the portfolio is still growing.
A method for suppressing hallucination in language models at the moment of inference, before an answer becomes language.
A constraint layer that operates on model logits during generation, asynchronously enforcing limits as tokens are produced rather than filtering output after the fact. Architecture, claims, and reference behaviour are set out in the filing.
Hallucination is the single largest barrier to deploying language models in regulated and high-stakes settings. Buyers carry the liability for confident wrong answers. Post-hoc filtering is slow, partial, and easy to route around.
Filed and available for license. Acting before enforcement tightens secures the method at the layer every deployer will need, rather than negotiating for it once it is unavoidable.
An architecture for machine reasoning that responds before it deliberates, separating reflex from deliberation in the control path.
A reasoning architecture that routes between fast reflexive responses and slower deliberative paths, designed to span neuromorphic and quantum-classical substrates. The control structure and claims are set out in the filing.
Inference cost and latency scale with every deliberative step. Buyers pay for compute they do not always need. A reflex path that answers the easy cases cheaply addresses both bills and response time.
Filed and available for license. Securing the routing layer early means owning it across the hardware transition, not competing for it after the substrate has settled.
A quantum approach to allocating capital under uncertainty, integrating topological methods into the optimization itself.
An optimization method that incorporates Fibonacci anyon structures into portfolio allocation under uncertainty, designed for quantum and hybrid execution. The formulation and claims are set out in the filing.
Allocation under uncertainty is the core problem of quantitative finance. Marginal improvements in optimization translate directly into return and risk. Institutions invest heavily to hold any durable edge.
Filed and available for license. Holding the method before quantum execution is commonplace positions the buyer ahead of the curve, not chasing it.
A method for molecular and pharmaceutical discovery with auditability built into the discovery process rather than added after it.
A platform that fuses neuromorphic computation with temporal-logic programming, so that each step in molecular synthesis carries its own explanation and audit trail. The architecture and claims are set out in the filing.
Pharmaceutical discovery is bound by regulatory evidence. Buyers spend years reconstructing why a result holds. A method that is auditable by construction reduces the gap between discovery and approval.
Filed and available for license. Owning an audit-native method ahead of regulatory tightening turns a future obligation into a present advantage.
Further methods are in drafting now and moving toward filing, on the same thesis: making scarce compute behave with intent. Like everything here, they are conceived and held by the founders of Nakatomi. Earlier conversations tend to be the better ones.
Licensing is the product.
Each method is filed and available for license. Customers build the interface; Nakatomi licenses the control surface. A protected method has to be negotiated with.
This is a founder-led company in the literal sense. Every method, filed or in drafting, is owned outright by the founders of Nakatomi, with no investors, third parties, or encumbrances on the title. You negotiate with the people who built it, and what is agreed is what holds.