Our Technology
Xyplor is built on proprietary technology covered by five US provisional patent applications filed in 2026. Below are plain-English descriptions of what each one does. They cover the systems that let kids finish ambitious work, return polished results, run reliably, keep generative AI safe for a child, and stay economical at scale.
Guided long-form creation
A system that lets a child carry one ambitious project across many sessions — recognizing when each milestone is genuinely complete and advancing the project automatically — so kids finish real, substantial work instead of abandoning it halfway.
Self-improving creations
A feedback loop that reviews a child's in-progress creation against quality criteria and re-prompts the AI to raise the bar before the result is shown, so first drafts come back closer to finished — without the child having to know how to ask for the fix.
Reliable AI-authored games
A structured way of representing kid-built games, with inference that fills in what a young creator didn't think to specify and a multi-stage validation pass that catches broken or unplayable output before it reaches the child — so what a kid asks for actually runs.
Safe kid-facing mentor AI
A layered safety system that screens every interaction a child has with the AI mentor, with quiet escalation to a parent when something genuinely needs adult attention — the architecture that makes it responsible to hand generative AI to a six-year-old.
Predictable AI economics
A spend-governance system that tracks each family's AI usage against a budget, with tiered alerts and a fast in-memory accounting layer that avoids routing every request through the database — the plumbing that keeps Xyplor's per-family AI cost predictable as the platform scales, without slowing down any single creation.
Frequently asked
No. They are US provisional patent applications filed in 2026. Provisional applications establish a priority date with the US Patent and Trademark Office; full (non-provisional) applications are filed within 12 months. Nothing on this page should be construed as granted patent rights or as a specific scope of legal protection.
Both. The patent applications protect the specific structural mechanisms we believe are novel and would otherwise be reinvented by competitors. Trade secrets protect implementation details that don't benefit from public disclosure. The combination gives a young company a defensible posture without locking everything behind opaque code.
No. The patented mechanisms are architecture and reliability choices that, in aggregate, make Xyplor more dependable and cheaper to operate at scale. Patent filings are a one-time legal cost, not an ongoing licensing burden.
It means Xyplor isn't a generic AI wrapper. The systems that keep AI safe for kids, fair to parents, and predictable for school budgets are first-party engineered. For procurement reviewers, it's evidence that Xyplor invests in the substrate of the product, not just the surface.
We will not assert these patents against good-faith open-source projects, individual researchers, students, or non-commercial educators. Our intent is defensive — to prevent commercial competitors from copying the specific mechanisms we invested in disclosing — not to suppress derivative learning or open research.
Legal note. The five filings referenced on this page are US provisional patent applications, not granted patents. "Patent pending" indicates that an application has been filed with the USPTO; it does not indicate the scope of any granted patent rights. Nothing on this page constitutes legal advice or a specific claim of enforceable patent protection.
For partnership inquiries (research collaborations, licensing discussions, or to verify a specific aspect of our IP for procurement due diligence), reach us at partnerships@xyplor.com.