01
Authorize
Should this AI-assisted change be allowed to move forward?
Patents
Patents, product boundaries, and the systems behind them.
My independent IP work focuses on the control layer around consequential AI-assisted actions: authorization, reproducible remediation, replay, and provenance.
Pending and filed applications are described only at a public-safe level. Official USPTO records control legal status and claim scope.
- 1 issued
- U.S. patent with public claim language
- 3 additional filings
- Pending/filed U.S. utility applications described at a public-safe level
- Working implementation
- ReplayGuard pairs the issued patent with a tested public reference system
Portfolio
Four systems, described at their actual stage
The useful question is not how many ideas sit in a filing portfolio. It is what each system is trying to control, where it sits in a workflow, and what has actually been built or issued today.
Issued
Deterministic Offline Code Remediation
U.S. Patent US 12,670,085 B1 · issued June 30, 2026
Verify a bounded remediation before downstream authority is processed
The issued workflow applies template-based remediation in two runs, compares the processed code byte for byte, processes approval on a match, and records cryptographic plus ledger-entry evidence.
Open flagship asset →Pending
Deterministic Code Authorization
Pending U.S. utility application
Decide whether an AI-assisted code change may advance
A deterministic control concept for placing policy and authorization between an AI-assisted change and a consequential downstream step such as merge, release, or deployment.
Filed
Decision Replay & Ledger-Verified Execution
Filed U.S. utility application
Preserve enough context to reconstruct an AI-influenced action
A system concept for retaining the inputs, execution context, and recorded artifacts needed to understand a consequential decision after the original moment has passed.
Filed
Cross-Modal Provenance
Filed U.S. utility application
Track origin and transformation across multimodal artifacts
A provenance concept for preserving lineage across AI-generated and AI-transformed text, code, documents, and images without exposing pending claim language or proprietary implementation rules.
System thesis
Four control questions around AI execution
Model output can be useful long before it is safe or appropriate to act on it. These systems address different points in that gap.
02
Verify remediation
Can the proposed fix be reproduced under the defined process?
03
Replay
Can the action be reconstructed after the fact?
04
Prove lineage
Can an artifact's origin and transformation history be traced?
Flagship issued asset
ReplayGuard
Evidence before automated remediation receives authority.
ReplayGuard turns the issued workflow into something a reviewer can inspect: a tested reference implementation, a bounded browser-local evaluator, reproducible fixtures, and clear limitations. It is the strongest example in the portfolio of an invention moving from filing to working product artifact.
- Patent
- US 12,670,085 B1
- Issued
- June 30, 2026
- Independent claims
- 1, 16, and 17
- Public implementation
- Tested reference system + synthetic evaluator
How I use IP
Patent strategy should sharpen the product, not replace it.
Define the product boundary
Start with the user, workflow, control point, and the artifact the system must create.
Keep stages distinct
Issued rights, active filings, working implementation, and future product possibilities are different things.
Commercialize from working proof
Lead with technical fit, implementation behavior, integration surfaces, and explicit limitations before the broad commercial story.
Public, non-confidential portfolio summary. Official USPTO records and issued claim language control. Nothing on this page is legal advice, an infringement or validity opinion, or a binding commercial offer.