Remnant Fieldworks Inc.
Stop unverified AI actions
before they happen.
ExecutionProof checks whether an action is authorized, supported by evidence, and within policy before allowing it to execute.
Built for AI agents, automated workflows, payments, deployments, access changes, and other consequential actions.
Proof Before Power™ · Verification Before Execution™
The Core Shift
Why Verification Before Execution™?
Most systems act first and check later. We move the checkpoint to the moment that actually matters — before the action fires.
Without ExecutionProof
Executes first
The action fires before anyone confirms it should.
Audits later
Review happens after the fact — if at all.
Discovers failure afterward
Damage is found once it is already done.
With ExecutionProof™
Verifies first
Authority, policy, evidence, state, and risk are checked up front.
Executes second
Only verified actions are allowed to proceed.
Generates proof automatically
Every decision leaves a signed, tamper-evident record.
The Flagship Platform
ExecutionProof™
An independent pre-execution verification and control layer for consequential automated decisions. Before a consequential action is allowed to execute, it evaluates the actor, the exact proposed action, authority, evidence, constraints, policy, target, parameters, and current state. Payment authorization is the first production boundary. The same architecture governs deployments, access changes, data release, infrastructure commands, and any irreversible action.
Verification Gate
ExecutionProof™ receives a proposed action, verifies the actor, authority, evidence, policy, constraints, target, parameters, and current state, returns ALLOW / HOLD / DENY, and generates a signed, tamper-evident ProofRecord™, a verifiable receipt of the execution decision, before the action proceeds. ExecutionProof does not process payments, deploy code, or release data. It sits before the downstream system, whether a payment rail, deployment pipeline, access-control layer, or infrastructure controller, and determines whether a consequential action has earned the right to execute. The downstream system executes only after ExecutionProof returns ALLOW.
Where a boundary requires two independent approvals before release, ExecutionProof™ holds the action until valid, independent approval evidence is supplied — enforcing M-of-N thresholds and refusing self-approval — then re-verifies before release. Approval routing, notification delivery, and identity integration are configured per pilot.
The flagship verification engine is operational today — a live /v2/verify endpoint returning real ALLOW / HOLD / DENY decisions with signed ProofRecords™ and working HOLD resolution. ExecutionProof implements hardware-backed ML-DSA-65 post-quantum signing for ProofRecords through AWS KMS. Formal RF-100 §8.4 conformance remains pending independent external review. Live pilot integrations are open now.
ExecutionProof™ is a Remnant Fieldworks™ platform built on the Proof Before Power™ doctrine and the Verification Before Execution™ framework. Each decision produces a ProofRecord™ for auditability, reconstruction, and accountability.
Explore ExecutionProof.ioAn Engineering Platform
Build Against the Boundary
ExecutionProof is something you build and test against, not a whitepaper you read. These are the things engineering and security teams can do with the boundary during a pilot integration.
Try to bypass the execution gate
Send proposed actions straight at the boundary and attempt to get an unauthorized action through without a valid ALLOW.
Test delegated authority between AI agents
Chain agents, pass authority down a delegation path, and probe whether the boundary still enforces the original grant.
Detect conflicting policies
Load competing or overlapping policies and check whether the gate resolves the conflict or lets an unsafe action slip.
Run conformance cases across implementations
Exercise the RF-100 verification behaviors as repeatable conformance cases and compare decisions across implementations.
Integrate one real workflow through /verify
Route one real, consequential workflow through the verification endpoint and watch ALLOW / HOLD / DENY decisions in the loop.
Want boundary access?
Verification endpoint access and conformance cases are provisioned per pilot integration.
Start with One BoundaryOpen Invitation
Challenge ExecutionProof
We want engineers, researchers, students, and security teams to try to break the boundary. Build adversarial harnesses, test delegation paths, identify policy conflicts, reproduce results, and tell us where the architecture fails.
Build adversarial harnesses that hammer the execution gate.
Test delegation paths where authority passes between agents.
Identify policy conflicts the boundary should catch.
Reproduce our published results and check the claims.
Findings that expose a real weakness make the boundary stronger. Tell us what you tried, what you observed, and where it broke.
Where It Applies
One Boundary, Many Consequential Actions
The same execution gate — verify authority, evidence, and policy, then return ALLOW, HOLD, or DENY — applies wherever an unverified action carries real-world consequence. Agent-initiated payment is the first production boundary; these are the domains the same architecture is built to govern.
AI Agents & Autonomous Systems
Intercept and verify AI-generated actions before they execute — ensuring authority, intent, and policy compliance at the moment an agent acts, not after.
Enterprise Automation
Apply proof-gated execution control to automated workflows, RPA, and orchestration layers — so no high-consequence process runs on cached permission alone.
Data Centers & AI Compute
Inheritance-bound governance for power allocation, cooling, grid capacity, and infrastructure provisioning — verification of physical-domain conditions before compute executes.
Treasury & Payments
Proof-gated authorization for capital movement, wire transfers, payment rails, and settlement — verifying authority and limits before money moves.
Digital Assets & DeFi
Pre-execution verification for wallet operations, smart contract interactions, and on-chain transactions — governance that intervenes before irreversible transfers.
Insurance & Risk
Enforce underwriting constraints at the execution boundary — ensuring coverage conditions, exclusions, and authority are verified before claims or policy actions proceed.
Healthcare & Regulated Systems
Verify authorization, safety constraints, and regulatory compliance before care actions, prescriptions, or regulated processes execute.
Critical Infrastructure
Hold high-impact commands — grid operations, industrial control, safety-critical systems — until every condition, constraint, and authority is verified and recorded.
8 pending non-provisional U.S. patent applications · Patent pending
Public Research Record
106 Documented Experiments Across a Growing Public Research Record
Remnant Fieldworks uses a design-before-execution research discipline: define the question, preregister where applicable, preserve negative results, diagnose failures, separate remediation from the original result, and publish the record. Every claim behind ExecutionProof™ is backed by a preregistered, publicly archived experiment corpus, including the results that did not pass. What began as a single 18-experiment ARK cycle is now a reconciled research program spanning six phases and twelve research families, including preserved negative results, adversarial security testing, IBM Quantum hardware studies, intent-fidelity testing, and execution-boundary research. The record is designed to be audited, not admired.
“Permission at approval time is not necessarily permission at execution time. Verification must occur at the execution boundary.”
106 documented design-before-execution experiments across 16 public repositories: 95 PASS, 8 preserved FAIL, 3 special status, with 4 validated FAIL→PASS remediations. 90 published Zenodo works across 124 versions under open-access licenses. The eight preserved FAILs: ARK-445, ARK-455, DM-001, QG-001, QG-002, EP-SEC-009, IB-001, and IB-002. Four were closed by separately preregistered remediations, with the originals preserved rather than overwritten: ARK-445b, ARK-455b, EP-SEC-009b (a call-origin guard closing a direct tool-invocation bypass), and IB-003 (restoring intent integrity after IB-001 and IB-002). DM-001 (classical baseline), QG-001 and QG-002 (IBM hardware noise) remain unremediated and honestly preserved. Three special-status records: ARK-448 GATE-STOP, ARK-502 SMOKE-PASS, ARK-503 NOT-EXECUTED (awaiting independent review).
The eight preserved FAILs and three special-status records are not gaps in the record; they are the record. A corpus that only ever reports success is not evidence; it is marketing. Internal and founder-led research. Independent academic and external validation remain the next phase.
Corpus by phase
The full ExecutionProof experimental record, organized by phase from hardware authorization boundary through enterprise readiness.
Hardware authorization boundary
8 IDs · 7 PASS · 1 FAIL · 1 GATE-STOP
The founding ARK cycle — tested on real IBM Quantum hardware. ARK-445 FAIL (preserved) → 445b retest PASS. ARK-448 GATE-STOP (preserved). Current-state authority, exact-action binding, workflow isolation, three-state ALLOW / HOLD / DENY, and self-approval refusal.
Synthetic boundary completion
9 IDs · all PASS except ARK-455 FAIL (preserved)
ARK-455 FAIL (preserved) → 455b retest PASS. Synthetic phase CLOSED. ARK-457 published (DOI 10.5281/zenodo.21421742). State-change attacks, substitution, revocation, workflow, conflicting-evidence, self-approval, dependency-loss, and cross-context replay patterns addressed.
Production-boundary matrix
25 IDs · 25 / 25 PASS · 20,000 decisions
5 real-world domains (cloud IAM, production deployment, destructive DB action, financial transaction, API policy/rate limits) × 5 failure modes. 20,000 / 20,000 dual-guard agreement, 5,306 / 5,306 wrong-allows detected. Bounded test environments — not production certification.
Latency / throughput / scale
10 IDs · all PASS · all published
Verification p95 1.822 µs (Python) / 0.652 µs (JS); burst 1.66 M/s (Py) / 4.52 M/s (JS). Bounded single-threaded in-memory engineering evidence — not production certification. Includes preserved public correction to the ARK-486 cost model. DOIs 10.5281/zenodo.21434398–21434413.
Enforcement-boundary integrity
6 IDs · 161 / 161 PASS · 0 enforcement leaks
Mechanical enforcement-boundary integrity: the gate holds across execution paths with zero enforcement leaks under adversarial load, deep mutation, timing races, and delegation abuse. Includes 10 / 10 independent tamper detection with 0 false positives (ARK-497) and fail-closed behavior with 100% independently signature-verifiable proofs in a production-like networked run (ARK-498).
Enterprise adapter & operational readiness
5 IDs · 499/500/501 PASS · 502 SMOKE-PASS · 503 awaiting review
ARK-499/500/501 EXPERIMENT-PASS (7/7 each; real self-hosted PostgreSQL 17, git CI/CD, RS256 OIDC/JWKS boundaries). Core finding: the authorization boundary did not need to change to meet real enterprise surfaces — only the adapters did. ARK-502 SMOKE-PASS (418 ops, 0 leaks; ≥14-day endurance NOT executed, 0 scored). ARK-503 NOT-EXECUTED (reviewer package delivered, awaiting human reviewer, 0 scored). Labels: real but self-hosted (NOT Docker/K8s/cloud; NOT Okta/Azure AD/Auth0). Post-quantum custody milestone: ML-DSA-65 signing key now held in AWS KMS FIPS 140-3 Level 3 HSM — hardware-backed, not in-process. RF-100 §8.4 conformance pending independent external review only.
Quantum-evidence & scientific research
Quantum-sourced evidence, intent-fidelity testing, intent binding, authority partitioning, adversarial security, dark-matter simulation, and broader coherence research. Real-hardware experimental evidence, not new physics, not universal security proofs. Device-dependent, not loophole-free. Includes preserved FAILs from hardware noise and adversarial testing.
Quantum-sourced authorization evidence
WITNESS-3: CHSH S = 2.545 (15.8σ)*
Provider-record verification, tamper / substitution detection, replay rejection, and multi-source nonce construction anchored to IBM quantum measurements, NIST randomness-beacon data, and LIGO / GW150914 archival data. WITNESS-4 is planned and excluded from the corpus count — do not cite as executed.
Quantum-hardware correlations
S = 2.514 · |M| = 3.423 · χ = 5.268–5.376*
Bell-CHSH spatial correlations, multipartite Mermin correlations, and Peres–Mermin contextuality measured on quantum hardware.
Temporal witness
CHRONO-1: K3 = 1.450*
A Leggett–Garg temporal-inequality test bound into an authorization record, under the documented device conditions.
Combined witness in one governed job
S = 2.797 · K3 = 1.502 · χmin = 5.126 · 52,000 shots · ALLOW*
Spatial Bell-CHSH, temporal Leggett–Garg, and contextual Peres–Mermin witnesses fused into one governed hardware job. A published six-point erratum (v1.1) corrects the shot-count, nonce, contextual-significance, K3 wording, novelty wording, and a qubit-mapping logging defect.
Three-processor fused witness
TRINITY-1: S = 2.736 / 2.579 / 2.741 across 3 devices · 48,000 shots · ALLOW*
The same Bell-CHSH witness independently evaluated on three distinct IBM Heron r2 processors (ibm_kingston, ibm_fez, ibm_marrakesh) — no device averaged away. Three device results are cryptographically fused into one chain-linked, independently reconstructable ExecutionProof record. TRINITY-2 extends this to a governed simulated payment authorization with quorum logic.
Dark-matter quantum simulation
DM 001–005: 4 PASS · 1 FAIL (DM-001)
Preregistered quantum simulations of dark-matter candidate models (axion, sterile neutrino, WIMP) using IBM Q hardware. DM-001 FAIL (classical baseline, unremediated) is honestly preserved. Toy Hamiltonians only — not detection of real dark matter.
Pre-boundary intent fidelity
IF-01–IF-06: 6 / 6 PASS
Measuring model-introduced divergence between authenticated request and proposed action before the ExecutionProof boundary. Tests whether intent survives the model layer without paternalistic distortion.
Quantum governance
QG 001–004: 2 PASS · 2 FAIL (QG-001, QG-002)
Can evidence from imperfect quantum hardware be trusted enough to cross an execution boundary? QG-001 and QG-002 FAIL (IBM hardware noise, unremediated) are honestly preserved.
Adversarial security testing
EP-SEC 001–009b: 9 PASS · 1 FAIL (EP-SEC-009)
Security boundary testing against adversarial attack vectors. EP-SEC-009 FAIL (direct tool invocation bypassing the boundary) is a genuine adversarial finding and remains preserved in the record. EP-SEC-009b is a separate, design-before-execution remediation retest: the side-effect path now enforces a call-origin guard, and all eight adversarial cases passed, including a positive control on the legitimate path. The original FAIL is not retconned. EP-SEC depositions not yet on Zenodo.
Intent binding
IB-001–003: 1 PASS · 2 FAIL (IB-001, IB-002)
Does an authenticated request’s intent survive binding into an authorization record? IB-001 and IB-002 FAIL are honestly preserved; IB-003 is the separately preregistered remediation that passed all twelve cases. The intent-integrity layer of the consolidated intent-binding / authority-partitioning research cluster.
Authority partitioning
AUTH-001–002: 2 / 2 PASS
Authority integrity and composition safety: can authority be partitioned so that no single actor — and no composed action — exceeds its granted scope? AUTH-001 (authority partitioning) and AUTH-002 (composition safety) each passed all eighteen preregistered cases.
Coherent Inheritance Framework
The Coherent Inheritance Framework is an ongoing research framework examining how meaning, confidence, validity, contradiction, and provenance survive across transformations and handoffs.
* Quantum results hold within their stated device-dependent conditions and tested circuit model. They are demonstrations of quantum-sourced evidence and provenance, not general physical claims beyond those conditions.
How the record is kept honest
Preregistration locks
Hypotheses and a SHA-256 MANIFEST are committed before execution — criteria cannot move after results are known.
Fixed pass/fail criteria
Success conditions are defined up front. No post-hoc redefinition of what counts as a pass.
Dual independent verifiers
Records are checked by independent verification paths before entry into the corpus.
Published corrections
Failures, gate-stops, and errata are published in full alongside successes — disclosed through a formal public correction process, never quietly removed.
Git provenance
Public commit history with MANIFEST SHA-256 hashes provides a timestamped, tamper-evident record of each experiment.
Live archival DOIs
Each series is archived on Zenodo with permanent concept DOIs — citable, versioned, and open for external critique.
Source repositories
Remnant Fieldworks maintains sixteen public repositories: fifteen experimental, implementation, and research repositories, plus the separate RF-100 standards repository.
executionproof-testbeds
Canonical ARK / P01 / P02 track
witness-testbeds
WITNESS quantum-evidence series
bellwether-testbeds
BELLWETHER correlation series
chrono-testbeds
CHRONO temporal-witness series
omni-testbeds
OMNI combined-witness series
cif-phase1-testbeds
Coherent Inheritance Framework — Phase 1
trinity-testbeds
TRINITY cross-processor fused witness series
vaultproof-agent-guard
VaultProof agent-wallet guard
cif-ancient-systems-test-series
CIF ancient-systems test series
dark-matter-quantum-sim
Quantum simulation of dark-matter candidate models on IBM Q, preregistered with locked SHA-256 manifest
inheritance-math
CIF mathematical modeling of knowledge, authority, and meaning across boundaries
executionproof
ExecutionProof pre-execution governance layer
intent-fidelity-testbed
FIDELITY intent-fidelity testing series
quantum-governance-testbed
QG quantum governance experiments
rf-100
RF-100 standard · separate standards repository
Archived & citable
Research series (concept DOI)
90 published Zenodo works across 124 versions under open-access licenses, including series concept records and their versions (EP-SEC not yet deposited). The concept DOI always resolves to the latest archived version of the series.
DOI 10.5281/zenodo.21398675OMNI-1 — cite the concept DOI
A six-point erratum (v1.1) corrects the original OMNI-1 record. Cite the concept DOI, which resolves to the corrected version; original and corrected versions are both preserved.
Research program, separately labeled
CIF-LAAD Validation Series
What information should still deserve trust when observations become incomplete, contradictory, stale, or uncertain? CIF-LAAD, the Coherent Inheritance Framework for Low-Altitude Air Domain Awareness, is a simulation-stage research program exploring continuity, false-track suppression, uncertainty, evidence provenance, and computational behavior in low-altitude sensing environments. It is a Remnant Fieldworks research program and controlled implementation, not a replacement for ExecutionProof. CIF-LAAD is being investigated as a continuity-and-assurance layer for low-altitude sensing, particularly when observations disappear, conflict, or become cluttered. It is research only.
Counting rule: The CIF-LAAD Validation Series is a separate, preregistered study. Its held-out seeds are replications, not new experiments, and it does not change the 106-experiment corpus count above.
What the evidence supports
- Full simulated continuity through about 5 seconds of total sensor dropout within the tested envelope.
- Strong reduction in false-track persistence under the tested clutter and crossing conditions.
- Bounded identity-switch improvement at lower target densities.
- Tamper-evident evidence-chain verification.
- Typed rejection of hostile or invalid observations.
- Reproducible, preregistered validation with held-out seeds and frozen engine hashes.
Documented limits
- External JPDA and GNN comparators achieved better positional RMSE in most tested scenarios.
- Localization accuracy is materially worse than baseline under contested scenarios.
- Identity stability is not superior at high target densities.
- Plausible schema-valid spoof observations are not automatically rejected.
- Real-time operation is bounded to roughly 50 tracks in the tested pure-Python configuration.
- No certified third-party tracking benchmark yet.
- No hardware testing and no independent validation.
- The preregistered coherence-gated inheritance hypothesis was falsified and is preserved as such.
CIF-LAAD and ExecutionProof answer different questions
CIF-LAAD research asks what evidence or state may still deserve trust. ExecutionProof asks whether a consequential downstream action is authorized to occur. The two programs remain technically and evidentially independent.
Sensors and evidence → CIF-LAAD (trust, continuity, provenance) → customer C2 or mission system → ExecutionProof where applicable (authority, evidence, policy, constraints) → ALLOW / HOLD / DENY → execution boundary.
CIF-LAAD stops at evidence and C2 handoff. It performs no weapon authorization, no firing control, no engagement optimization, no jamming, and no autonomous defeat logic.
How the architecture fits together
The Standards Pathway
RF-100: Emerging Field Standard
RF-100 is Remnant Fieldworks’ emerging field standard for verification-first execution control. It is being developed as a practical conformance pathway for organizations that need proof before high-impact execution.
What RF-100 Defines
RF-100 defines the minimum structure a high-impact system should satisfy before execution:
RF Standard Architecture
Research & Public Record
RF-100 is supported by a preregistered experimental corpus of 106 experiments (95 PASS, 8 preserved FAIL, 3 special status) across 16 public repositories — including the ARK enforcement-boundary and enterprise-adapter series, the OMNI, TRINITY, WITNESS, BELLWETHER, and CHRONO quantum-evidence series, dark-matter simulation, intent-fidelity testing, quantum governance, and adversarial security testing — all archived, citable, and open for critique.
RF-100 Draft (archived, citable)
DOI 10.5281/zenodo.21366342Open review repository
github.com/derekhone/rf-100Public critique invited; feedback incorporated with attribution.
Research discipline
Preregistered hypotheses with published negative results — failures reported in full alongside successes, before any commercial claim.
Innovation & Intellectual Property
A Patent-Pending Governance Architecture
Remnant Fieldworks™ maintains 8 pending non-provisional U.S. patent applications (1 parent and 7 continuations-in-part), directed to proof-first, pre-execution governance for AI, automated, human-authorized, financial, infrastructure, privacy, consent, ownership, insurance, and other high-consequence systems.
Patent-Pending Portfolio
8 Pending Applications
Eight pending non-provisional U.S. utility applications: one parent and seven continuations-in-part. All commercially significant subject matter is consolidated into this non-provisional stack.
App. Nos. 19/529,283 · 19/731,050 · 19/731,090 · 19/731,110 · 19/731,118 · 19/732,539 · 19/732,564 · 19/732,615 · Patent Pending
RFCA
Certification Authority
The Remnant Fieldworks Certification Authority — the certification pathway for verifying RF-100 conformance.
Under DevelopmentConformance Profiles
Domain-specific RF-100 profiles tailored to the realities of each high-impact vertical.
Non-Provisional Portfolio
Eight pending U.S. non-provisional utility applications: one parent and seven continuations-in-part.
| App. No. | Filed | Portfolio Name | Scope |
|---|---|---|---|
| 19/529,283 | Feb 4, 2026 | Parent / Foundational | Policy-gated execution with proof-first authorization and tamper-evident evidence artifacts; anchors the execution-bound governance spine and priority continuity from the earlier provisional filings. |
| 19/731,050 | Jul 4, 2026 | AI Compute Infrastructure | Data center / compute infrastructure governance and readiness for AI execution environments (power, cooling, water, grid, inheritance burden vectors). |
| 19/731,090 | Jul 5, 2026 | AI-Generated Actions | Governance of AI-generated actions: authority, evidence, constraints, execution admissibility; agent gating, M-of-N approval, self-approval blocking. |
| 19/731,110 | Jul 5, 2026 | Tamper-Evident Proof | ProofRecord™ / audit-clean evidence generation for execution events and verifiable runtime outcomes. |
| 19/731,118 | Jul 5, 2026 | Financial Execution | Financial execution governance, authorization, verification, and proof around high-impact transaction and treasury events. |
| 19/732,539 | Jul 6, 2026 | AI Safety / Consent / Privacy | AI safety monitoring, consent-based access control, privacy-preserving data governance in autonomous computing environments. |
| 19/732,564 | Jul 7, 2026 | Stewardship / Ownership | Ownership-bound execution control, stewardship verification, field-interlock access management. |
| 19/732,615 | Jul 7, 2026 | Risk / Insurance / Claims | Risk-adaptive governance, coverage-bound execution control, verifiable claims processing for AI and autonomous systems. |
Strategic Portfolio Map
One patent family, organized into five strategic pillars — from the foundational execution spine to the domains it governs.
Foundation
The execution-bound governance spine.
Infrastructure
Compute readiness, safety, consent, privacy.
AI Governance
AI actions, ownership-bound control, stewardship, access.
Proof / Audit
Tamper-evident proof of execution events.
Financial / Claims Execution
Money movement, risk-adaptive coverage, claims.
Filing Timeline
Provisional priority foundation
Priority foundation spanning the full pre-execution governance architecture: proof-first authorization, runtime enforcement, evidence and audit, rights and consent, treasury, infrastructure, stewardship, insurance, and commercial governance.
Parent non-provisional, U.S. App. No. 19/529,283
Policy-gated execution with proof-first authorization and tamper-evident evidence artifacts; establishes the execution-bound governance spine and priority continuity from the earlier provisional filings.
7 continuations-in-part, 19/731,050 through 19/732,615
Continuation-in-part wave extending the spine across AI compute infrastructure, AI-generated actions, tamper-evident proof, financial execution, AI safety and consent, ownership and stewardship, and risk, insurance, and claims.
Architecture Themes
Applications pending before the USPTO. No assurance is given that any claim will be granted. This information is provided for general portfolio description and does not constitute legal advice.
“Patent pending” means U.S. patent applications have been filed and are pending examination. It does not mean any patent has issued, that any claim has been allowed, or that the USPTO has endorsed the technology. Patent rights, if any, are subject to examination and issued claim scope.
The Governing Doctrine
Proof Before Power™
Before systems are given power to act, they must be able to prove authority, evidence, state, and control. This is the doctrine behind everything Remnant Fieldworks builds.
RF-000.1 · Core Law
“If it cannot be verified, it cannot execute.”
Under RF-100's proposed requirements, every action carries a verifiable authorization chain before execution begins.
RF-000.2 · Core Doctrine
“Verification Before Execution™.”
The sequence is invariant: verify first, execute second. Verification is the gate. The sequence cannot be reversed, bypassed, or deferred.
RF-000.3 · Core Philosophy
“Proof Before Power.”
Authority must be proven, not assumed. Power over systems, capital, or people is legitimate only when the authorization behind it can be independently verified.
Control · Decision · Execution
“Capability does not equal authority; execution should be permitted only when authority, evidence, constraints, state, and admissibility are verified at the execution boundary.”
This doctrine is documented in depth across the founder’s published work — 9 published books and 75+ articles on pre-execution governance, the verification boundary, and proof-first authorization.
The Story in Five Words
Doctrine → Framework → Standard → Platform → Proof
Doctrine
The governing belief
Framework
How verification works
Standard
RF-100 conformance
Platform
ExecutionProof
Proof
The signed record
Ecosystem
One Architecture. Three Entry Points.
Remnant Fieldworks™ is the parent company and IP architecture layer. The platform and console are purpose-built entry points into the same proof-first governance stack.
ExecutionProof™
The Platform
The independent pre-execution governance layer for consequential AI actions. ExecutionProof sits before the execution boundary — verifying authority, evidence, constraints, and state before any downstream system is allowed to act. The architecture extends to any domain where unverified action carries real-world consequence.
"If it cannot be verified, it cannot execute."
Boundary Console
The Builder
The ExecutionProof Boundary Console. Define high-consequence actions, map authority rules, configure evidence and constraint requirements, simulate ALLOW / HOLD / DENY decisions, and generate Execution Boundary Profiles for pilot or integration.
"Define the boundary before the system acts."
All three properties share the same patent-pending governance architecture, doctrine, and standards framework. Each serves a distinct role in the Remnant Fieldworks™ ecosystem.
What We Offer Today
One Free Tool. Two Remote Paths.
Today, Remnant Fieldworks offers one free open-source reference tool and two remote commercial paths: a Validation Sprint and a Supervised Pilot. Hosted plans, RF-100 assessments, and certification pathways will come later as the standard, backend, and crosswalk mature.
The Sprint and Pilot are not additive. Because 100% of the Sprint fee is credited toward the Supervised Pilot, your total engagement cost is the final Pilot price — a typical complete Sprint-to-Pilot engagement is approximately $25,000 total.
VaultProof Agent Guard
Open-source experimental reference tool for AI-agent wallet governance.
- ALLOW / HOLD / DENY logic
- Local policy checks
- Sample profiles
- Spend caps
- Allowlists
- Drain-vector blocking
- Telegram HOLD approval path
- Testnet / disposable-wallet use only
Experimental open-source reference implementation. Not audited. Not for production funds.
Request AccessRemote Execution Boundary Validation Sprint
Define one real execution boundary before a pilot.
- One remote kickoff
- One workflow / boundary review
- One governed-action definition
- Authority / evidence / constraints / state mapping
- ALLOW / HOLD / DENY recommendation
- Execution Boundary Profile draft
- RF-100 alignment memo
- Pilot recommendation
- 100% of the Sprint fee credited toward a subsequent Supervised Pilot
Timeline: approximately 10 business days
InquireSupervised Pilot
Test ExecutionProof against one real workflow.
- One defined execution boundary
- Remote setup
- ALLOW / HOLD / DENY test cases
- ProofRecord / VDR output
- HOLD and DENY no-execution testing
- Replay / expiry / parameter-change testing where applicable
- Pilot findings report
- Production-hardening roadmap
- You pay only the remaining balance after the Sprint credit is applied
Weeks 1–2
Boundary setup, remote integration, and policy configuration
Weeks 3–4
ALLOW / HOLD / DENY test execution and ProofRecord capture
Weeks 5–6
Findings report and production-hardening roadmap
4–6 week engagement
InquireDesign-Partner Pricing
Design-partner pricing (Supervised Pilot from $15,000) is available only in exchange for written case-study / testimonial rights, a recognizable reference, or a defined expansion path.
Coming Soon · Waitlist Only
These paths open as the platform, hosted product, crosswalk, and certification pathway mature. Contact us to join the waitlist.
The ExecutionProof™ Series
A governance library of 9 published books and 75+ articles documenting the doctrine, framework, standard, and platform — the written backbone behind the company.
9
Books Published
75+
Articles
#1
Amazon Kindle · Free
The Company
Remnant Fieldworks™
The company behind ExecutionProof — the verification layer that checks whether a consequential action is authorized, evidenced, and within policy before it executes. Patent applications pending.
Derek Hone
Founder & Named Inventor
Sole named inventor across the company's 8 pending non-provisional U.S. patent applications (USPTO Customer No. 236398), filed pro se. Author of 9 published books and 75+ articles. Creator of the Proof Before Power™ doctrine, the Verification Before Execution™ framework, and the RF-100 governance standard.
Remnant Fieldworks began with a single thesis: high-consequence systems should not be allowed to execute merely because an agent, user, or workflow requested action. Every system in the portfolio — from AI governance to data-center operations to financial execution — enforces that thesis at the execution boundary.
Derek Hone
Founder
Adith Kadam Ramesh
Systems Engineer
8 Pending Non-Provisional
Patent Applications
9 Books · 75+ Articles
Published Works
RF-100
Governance Standard
Company Milestones
September 1, 2026
NSF PESOSE Track 3 Proposal Submitted
Remnant Fieldworks Inc. submitted NSF PESOSE Track 3 Proposal #2641427, "PESOSE: Track 3: ExecutionProof, An Open-Source Ecosystem for Verifiable Pre-Execution Authorization of Autonomous Actions." Requested funding is approximately $1.5 million over 24 months if awarded. This is a submitted proposal, not an award.
September 1, 2026
NSF SBIR/STTR Project Pitch Submitted
Remnant Fieldworks submitted an America's Seed Fund / NSF SBIR-STTR Project Pitch in Cybersecurity and Authentication, Submission #00124422. This is a Project Pitch at the screening stage. It does not imply a Phase I invitation or award.
2026
Ohio Tech Day Participation
Remnant Fieldworks has submitted to participate as a 2026 Ohio Tech Day partner, committing to host an event, in person or virtual, promoted through company channels. The planned educational session is "Proof Before Power: How Do We Make AI Safer Before It Acts?"
Federal Snapshot - September 1, 2026
A dated, frozen record of the company's public research and intellectual-property posture at the time of its NSF proposal submission. These figures are historical. They are preserved as of that date and are not updated to current numbers. For the current live research record, see the Public Research Record.
On this date the company submitted NSF PESOSE Track 3 Proposal #2641427 and an NSF SBIR/STTR Project Pitch, Submission #00124422.
Academic proposal partners
The Ohio State University (Carter Yagemann) and the University of Dayton (Phu Phung, Luan Nguyen) are academic proposal partners and research collaborators. Their formal-analysis and security-evaluation roles are planned through the proposed NSF work, with future independent stress-testing and evaluation. This does not represent independent validation of Remnant Fieldworks technology.
Pilot Pathway
Start with One Boundary.
You don't have to govern everything at once. Pick one high-impact action, put it behind ExecutionProof, and prove the model on a single boundary.