Issued U.S. patentUS 12,670,085 B1Issued June 30, 2026

ReplayGuard · issued patent + tested reference implementation

Evidence before automated remediation receives authority.

ReplayGuard is a control layer around automated code remediation. It runs a bounded remediation twice, compares the resulting bytes, records cryptographic evidence, and routes the outcome through an explicit gate before downstream action.

Run public evaluation Read 1-page brief Watch 75-second overview

Problem

Automation can move faster than trust is established.

Generated or suggested patches can approach merge or deployment before repeatability has been checked.

Issued workflow

Two runs. Byte-level comparison. Recorded proof.

The issued workflow connects template-based remediation, repeated execution, matching, approval processing, hashing, and ledger evidence.

Product fit

A control point around systems that generate or govern fixes.

Potential integration surfaces include AppSec, AI coding agents, DevSecOps, CI/CD, and internal developer platforms.

How it fits

The control point between a proposed fix and downstream authority

ReplayGuard can sit around products that generate, suggest, apply, or govern automated remediation. The issued claims—not this product diagram—control legal scope.

Upstream

Proposed remediation

A scanner, AI coding agent, developer tool, or internal platform proposes a software change.

Code plus a remediation request
The upstream tool remains the detector or generator

ReplayGuard

Two-run verification

Apply the approved remediation path twice, compare exact output bytes, and create the required record.

First run + second run
Byte-level comparison
Hash + ledger-entry evidence

Downstream

Process authority

On the matched path, process the approval request associated with the claim-enumerated downstream actions.

Commit · promotion · patch
Merge · deployment · ledger update

Mismatch path

Dependent claims 5–6 · prototype BLOCK

Rollback and blocking are dependent-claim features; BLOCK is the reference implementation's product label.

No-template path

Dependent claim 9 · prototype REVIEW

No-template handling is separately identified; REVIEW is the reference implementation's human-routing choice.

AppSec remediation

Place replay verification between a proposed fix and the authority to merge or deploy it.

AI coding agents

Require repeatable remediation output before an agent-created change advances.

DevSecOps and CI/CD

Expose comparison results to policy checks, approval workflows, and review routing.

Developer platforms

Add a reusable control point around internal systems that create or govern software changes.

Product adjacency does not establish infringement. This diagram is not a claim chart, legal opinion, or representation of production deployment.

Browser-computed evaluation

Test the evidence gate with bounded synthetic scenarios

Guided mode covers the three core outcomes. Technical mode exposes six bounded synthetic cases with runtime evidence, full SHA-256 values, exact UTF-8 byte comparison, and downloadable synthetic receipts.

Evaluation Sandbox V2.1

Browser-computed

Six bounded synthetic cases with browser-computed SHA-256 and exact UTF-8 byte comparison.

Synthetic fixtures only. No uploads or untrusted execution.

1. Run the guided evaluation

Step 1 of 3

Follow a matching replay, a controlled mismatch, and an unsupported path in sequence.

01Matching replay
02Controlled mismatch
03Unsupported path

Matching replay

Credential remediation

Ready to compute

Two runs apply the same approved template and produce byte-identical output.

Fixture

Hardcoded credential

Runtime check

Run twice, compare exact bytes, and verify the configured evidence checks

View bounded scenario details

Finding: Hardcoded credential
Rule: RG-CRED-001
Requested action: Merge approval
Boundary: Independent-claim workflow illustration

2. Inspect runtime evidence

Nothing pre-approved

FixtureReady

Run twicePending

ComparePending

Evidence checkPending

DecisionPending

Synthetic input

const apiKey = "demo-secret";

Run 1 output

Awaiting execution

Run 2 output

Awaiting execution

Run 1 SHA-256

—

Run 2 SHA-256

—

Byte comparison

—

Evidence receipt

—

Technical evidence and claim-boundary annotation

Run 1 template

No template evidence generated.

Run 2 template

No template evidence generated.

Synthetic record SHA-256

—

Unkeyed hash of the generated synthetic receipt; not a digital signature or an immutable-ledger claim.

Independent-claim workflow illustration

The sandbox illustrates two template-based runs, byte comparison, approval processing, hashing, and evidence recording. The issued claims control legal scope.

What this run will establish

Whether two browser-generated outputs match exactly
Whether approved-template identity remains consistent
Whether the configured evidence check passes
Which prototype gate decision follows

Interpretation boundary

Exact replay establishes reproducibility under the selected synthetic fixture. It does not establish semantic correctness, universal security, or production readiness.

Run the current case to reveal the decision and downloadable evidence receipt.

Public synthetic evaluation—not a production enforcement platform. No uploaded or untrusted code is executed. Matching replay demonstrates bounded reproducibility, not semantic correctness, complete security, formal proof, customer adoption, or production CI/CD enforcement.

Loading the browser-computed evaluation sandbox…

Issued-patent facts

The patent is the controlling legal-technical record

The official patent PDF and its claim language control. The material below is a non-legal product explanation of the issued workflow.

Title: Deterministic Offline Code Remediation with Ledger-Verified Replay and Template-Based Patch Generation
Patent: US 12,670,085 B1
Application: 19/545,244 · filed Feb. 20, 2026
Inventor / applicant: Apurv Gaurav

Independent claim forms

CLAIM 1

Method

CLAIM 16

Computer-readable medium

CLAIM 17

Computing system

Claim 1 workflow · plain-English product summary

Receive code

Receive computer-processable code, including visual or textual information or a building block.

Execute two runs

Select and apply a remediation template in a first run and a second run.

Compare bytes

Determine whether the processed code from the two runs matches byte for byte.

Process authority

On a match, process an approval request tied to a claim-enumerated downstream action.

Record evidence

Create a cryptographic hash and record the claim-specified evidence in a ledger entry.

Selected dependent-claim features

Claim 5Rollback of first-run processed code when the two processed outputs do not match byte-for-byte

Claim 6Restore a previous known-good state and block commit or deployment after a byte mismatch

Claim 8Violation notification and modification prevention when the second-run template differs

Claim 9Violation notification and modification prevention when no template is found in either run

Claim 10Static, rule-based, non-generative analysis

Claim 12Offline storage and application of remediation template data or instructions

Claim 14Syntax, AST, taint, compliance, and template-assignment inputs

Claim 15Enforcement plus rollback behavior

This is not a claim chart, validity opinion, freedom-to-operate analysis, or infringement opinion. Independent-claim requirements, dependent-claim features, specification examples, and prototype enhancements are presented separately.

Implementation evidence

What the public implementation shows—and what it does not

The reference system is useful because it is inspectable. It is also intentionally bounded.

Pinned source

Tested prototype snapshot retained for qualified diligence

Verified

Automated backend verification completed at the pinned snapshot

Build passed

Reference frontend production build completed

Patent scope

Issued patent

Independent claims 1, 16, and 17
Template-based first and second remediation runs
Byte-level matching, approval processing, hashing, and ledger evidence
Selected dependent features shown above

Prototype evidence

Reference implementation

Synthetic ALLOW, BLOCK, and REVIEW paths
Browser-computed SHA-256 and exact UTF-8 comparison
Template identity, configured checks, and evidence receipt
Tests, examples, APIs, screenshots, and documentation

Not represented

Current limitations

No production deployment or customer-adoption claim
Not a complete SAST or taint-analysis platform
No formal verification or semantic-equivalence proof
No immutable ledger, digital signatures, or production CI/CD enforcement claim

75-second overview

See the product thesis, issued workflow, and implementation limits

A short orientation for product leaders, technical reviewers, IP teams, and commercialization discussions.

Caption-led overview · 75 seconds · no autoplay

Read the overview as text

ReplayGuard addresses the gap between a proposed automated remediation and the authority to let that change proceed. The issued workflow performs template-based remediation twice, compares the processed code byte for byte, and creates claim-specified evidence around the matched path.

The public implementation demonstrates bounded ALLOW, BLOCK, and REVIEW outcomes using synthetic fixtures. It is a reference implementation, not a production SAST platform, formal-verification system, or customer-deployed enforcement product.

Review materials

One place to inspect the product before a diligence call

Start with the brief, confirm the issued patent, inspect the source, and run the evaluator. Deeper claim mapping, ownership material, and transaction documentation remain controlled.

Recommended starting point

Executive brief

A one-page review of the product thesis, issued workflow, implementation boundary, and transaction paths.

Open executive brief 02

Controlling source

Official USPTO patent

Review the issued grant and claim language for US 12,670,085 B1.

Open official patent 03

Public source and tests

Reference implementation

Inspect the testable prototype, authorization invariant, CI workflow, documented limitations, and local reproduction instructions.

Open public repository 04

Browser-computed evidence

Public evaluation sandbox

Run guided outcomes or inspect six bounded cases with SHA-256, exact byte comparison, and downloadable synthetic evidence.

Run the evaluation 05

Forwardable overview

Strategic overview deck

A seven-page non-confidential review of the authorization gap, issued workflow, product adjacency, implementation limits, and transaction paths.

Open strategic deck 06

Qualified next step

Strategic and technical diligence

Request a focused discussion about strategic fit, product adjacency, evidence needs, and transaction structure.

Request diligence

Available after a qualified discussion

Deeper diligence remains controlled

The public page is intentionally non-confidential. The materials below can be provided for a defined strategic, technical, transaction, or IP-review purpose after recipient and scope qualification.

Pinned implementation source and reproduction instructions
Detailed test inventory and execution evidence
Claim-to-implementation technical mapping
Continuation and patent-family filing records
Commercialization brief and transaction scope
Ownership and transaction documentation
Technical-transfer and know-how scope

Request qualified diligence

A focused first discussion can confirm strategic relevance, intended review scope, and which controlled materials are appropriate next.

apurvgaurav@gmail.com

Open diligence email draft

Commercial options

Acquire the issued patent, license it, or diligence the broader implementation package.

The issued asset is US 12,670,085 B1. A qualified transaction may also address the related continuation application, source code, tests, templates, architecture, APIs, examples, documentation, and technical-transfer know-how.

The exact transaction scope is established only in written transaction documents. Third-party dependencies remain subject to their applicable licenses.

Available structures

Complete patent acquisition
Exclusive license
Non-exclusive license
Field-of-use license
Patent plus source code and implementation know-how

Commercial terms are available following a qualified discussion.

apurvgaurav@gmail.com

Open email draft

Related continuation application filed; details available in qualified diligence.

Non-confidential public overview. Not legal advice, a valuation, an infringement analysis, or a binding offer. The issued patent and written transaction documents control.