Osonwanne Group is now patent-pending — twice

Two ventures. Two U.S. provisional patent applications, both filed June 24, 2026. One applied-AI company building — and protecting — what it ships.

Explore our patents

See the ventures

This week marked a milestone for Osonwanne Group LLC: both of our flagship AI ventures are now patent-pending. We filed two separate U.S. provisional patent applications — one for CactusComply, our AI-native Arizona tax-compliance platform, and one for CactusQuant, our self-improving AI trading system. Both provisionals are held by Osonwanne Group LLC, and both were filed on the same day.

As an applied-AI company — an AI venture studio that builds its own products and advises other businesses — protecting the engineering behind our work is core to how we operate. Here is what we filed, at announcement level: what the systems do, not how they work.

Two ventures, two provisionals

CactusComply

AI-native Arizona Transaction Privilege Tax (TPT) compliance, built so a small-business owner never has to lose sleep over filing.

U.S. Provisional Patent Application No. 64/097,375 (filed June 24, 2026).

CactusQuant

A self-improving AI trading system whose real innovation is its safety architecture. It runs on a paper account; we make no performance claims.

U.S. Provisional Patent Application No. 64/098,125 (filed June 24, 2026).

What CactusComply’s filing covers

The provisional covers the engineering systems behind the product — what they accomplish for customers — not the underlying tax rules. Three of them, in plain terms:

A self-optimizing voice AI that checks its own work. Our outbound agent reconciles what it says it did against what actually happened in the system, catching and correcting AI hallucinations before they ever reach a customer — and it improves over time by learning only from verified outcomes.
Secure, tenant-isolated AI retrieval. When an AI answers questions over sensitive financial data, isolation is not optional. We built it into the retrieval layer and designed the system to keep each business’s records walled off, plus automated detection of anything that even tries to cross that line.
A self-healing data-integrity pipeline. Tax rates shift constantly across roughly 150 Arizona jurisdictions. Our pipeline versions every change, propagates it safely, and tells the difference between a real data error and a perfectly normal variation — so we alert on what matters and stay quiet on what doesn’t.

What CactusQuant’s filing covers

CactusQuant is building an AI that manages a systematic trading strategy and improves itself over time — retraining its own models, tuning its own settings, and learning from its own results. The hard part isn’t getting an AI to change itself. The hard part is letting it change itself with capital at risk while engineering it so it can never endanger the account.

The filing covers that safety machinery and the guardrails that govern it — not any proprietary formula, and not a promise of returns. The platform runs on a paper account, and we make no performance or returns claim. What we protected:

An AI that retrains itself — but has to earn the keys. A newly generated decision policy directs no capital until it has proven itself out-of-sample against the incumbent across multiple market regimes, with every failure path falling back to a known-safe baseline.
A self-tuner under supervision. A higher-order governor watches the self-tuner for over-reaction, oscillation, or drift, so the system can neither thrash itself nor quietly freeze.
Honest by construction, with humans in the loop. The system grades itself only on realized results, never optimistic backtests; routine adjustments apply automatically while anything that moves real risk awaits human approval — and every change is logged and reversible.

Patents and trade secrets: what we did and didn’t file

A deliberate note on what we didn’t file. Across both ventures, our curated rule bases, rate data, models, parameters, and datasets stay protected as trade secrets. We patented the engines and their guardrails — the systems that make each platform unique — while keeping the proprietary know-how under the hood. That mix of patents and trade secrets is how we build durable, defensible technology and retain ownership of the intellectual property our companies create.

Why this matters for an applied-AI company

Two filings in one day, across two very different domains — tax compliance and quantitative trading — reflect a single thesis: real engineering advances, protected as enforceable assets, are how an applied-AI company compounds value. It’s also how we approach AI consulting for other businesses — building systems worth protecting, not demos. CactusComply is currently raising a pre-seed round; CactusQuant is advancing its safety-gated architecture on a paper account.