Voltry Certification Methodology · VCM-1
Every claim should show its source, method, and limits.
Voltry turns a GPU's own records into signed, replayable evidence for the people who lend on it, insure it, and buy against it. The method is designed so a counterparty can check the result without taking Voltry's word for it.
Draft materials for diligence, not reliance. Select numerical thresholds and grading parameters remain subject to ratification following an upcoming experiment with a national laboratory in the Rocky Mountain region.
ONLY
In this section
Follow the method from evidence to verification.
The category
Monitoring serves the operator. Voltry makes selected evidence verifiable by counterparties.
A Voltry certificate is a signed, replayable statement about one device, bound to an identity whose strength is stated on its face. It carries measured condition, operating history, and the completeness of the evidence behind both.
No price. No lien search. No adjudication. No blended score.Evidence semantics
Four evidence classes, kept separate.
Every claim keeps its type, provenance, and decision weight wherever it appears.
Measured
A fact read from the hardware
ECC counters, retired pages, serial and device configuration
Derived
A deterministic, replayable function
Duty odometer, incident rate, coverage and threshold distance
Modeled
A statistical estimate, kept decision-inert
Always banded, calibration-stamped and never part of a grade
Declared
A party-signed assertion
Ownership, custody, facility assignment and reservation
Three independent coordinates
Evidence class, assurance, and coverage remain separate.
How was the record taken?
Coverage
C0 scanC1 functionalC2 continuous
A point-in-time baseline is not presented as a continuous history.
How difficult is the evidence to fabricate?
Assurance
A0 self-reportA1 deviceA2 platformA3 witness
The class on the record states where the trust boundary stops.
What does the evidence mean under an agreement?
Verdict
PassFailNot assessed
Policy evaluation is mechanical, versioned and independently replayable.
Condition assessment
Seven axes stay visible. The weakest required axis binds.
VCM-1 defines no weighted or compensatory composite. Once VGR-1 is ratified, a condition grade is computed as the minimum band across required axes 1 through 5. Missing required evidence produces UNGRADED, never an optimistic average.
MIN
A gate, not an average.
A strong axis cannot compensate for a weak one.
Condition axis Role Indicator
01Vendor serviceability gradable
02Memory repair trajectory gradable
03Operational incident rate gradable binds
04Operating-state stability gradable
05Fabric health gradable
06Evidence continuity sets record tier
07Reservation and configuration integrity declaration-aware
Record tierBRONZECondition gradePENDING VGR-1
Missing evidence
Missing evidence is never a pass.
Required signal ∅ Not Assessed
A counter that was never read is not zero. A machine that stopped reporting is not healthy. Silence stays visible as its own outcome rather than being converted into reassurance.
Required evidence exists and satisfies the accepted rule.
Admissible evidence affirmatively violates the accepted rule.
The evidence needed to evaluate the rule is absent or unreadable.
Third-party verification
A retained package can be verified without Voltry.
A relying party that retains the verification package can check integrity, provenance and specified derivations without Voltry being online, solvent or cooperative.
- 01
Check the signature
Bundle plus signer public key
- 02
Rebuild canonical bytes
RFC 8785 and SHA-384
- 03
Confirm ledger inclusion
Proof plus signed root
- 04
Replay the result
Inputs under stamped versions
Operational maturity · August 2026
Current implementation status.
VCM-1 separates implementation from exercise and commercial use. The current field program routinely produces A0/C0 artifacts. Higher assurance, continuous coverage and a letter grade remain gated by their stated release and ratification conditions.
439 signed bundles 40+ registry identities 0 external counterparties yet
Evidence and verification core
- Read-only C0 capture
- Signed bundles and schema
- Registry, API and inclusion proofs
- Offline verification
Assurance and collateral workflows
- Deterministic policy engine
- Independent second evaluator
- Audience-scoped reporting
- Vendor serviceability rules
Higher-assurance operation
- A1 to A3 assurance
- C2 resident collection
- VGR-1 grading rubric
- Independent witness lanes
Read the source material
Email yourself the short overview and full methodology.
Both documents are distributed as clearly marked drafts. Submit one work email and we will send private open and download links for both. Links expire after 24 hours.
The argument
Voltry Litepaper
The category, trust architecture, counterparty applications, deliberate limits and what exists today.
- Version
- Draft r11
- Updated
- 2026-08-12
- Length
- 11 pages
- File
- Voltry_Litepaper_DRAFT_r11.pdf
Governing document
VCM-1 Methodology
The normative evidence classes, assurance and coverage ladders, condition axes, refusals, verification and maturity register.
- Version
- v0.3 draft
- Updated
- 2026-08-12
- Length
- 37 pages
- File
- Voltry_VCM-1_v0.3_DRAFT.pdf
Research access
Email both drafts to yourself.
One address, one message, both documents. We will not ask for additional details here.
Email sent
Check your inbox at your address. The message contains separate open and download actions for both drafts. Links expire after 24 hours.
Draft materials for diligence, not reliance. Select numerical thresholds and grading parameters remain subject to ratification following an upcoming experiment with a national laboratory in the Rocky Mountain region.