Controlled - NDA Stage

ARK Verification Command Center - Evidence Memo v1.0

Proof-boundary summary for a local evidence visualization layer around silicon verification and review-compression artifacts.

DateMay 2026

SurfaceEvidence visualization layer

BoundaryNot proof generation or signoff

What The System Does

Ark Verification Command Center is an evidence organization layer for silicon verification and review-compression artifacts. It presents review surfaces for overview status, RTL Review Witness evidence, affected cone summaries, provenance, replay summaries, manifest lineage, and evidence review.

The system does not generate final verification truth. Its value is organizing bounded evidence, known limitations, manifest references, replay lineage, and escalation paths into a reviewable local dashboard.

Review Compression Evidence Card

Review Compression403 RTL lines -> 8 root-signal targets
Measured on the current RTL Review Witness evidence corpus.

Preserved EvidenceCone JSON artifacts, localized SAT escalation evidence, manifests, and human-review limitations.

Boundary: This is measured on current internal/public-corpus artifacts. It is review-compression evidence, not final verification closure, customer ROI, or a guarantee that excluded regions are safe.

Implemented Surfaces

OverviewPortfolio status and risk cards

RTL Review WitnessScenario table and localized SAT evidence

Cone ViewerRoot, upstream, downstream, expressions

Sequence EvidenceFormal ladder, synthesis, workload packet

Review Evidencebounded artifact summaries and limitations

Evidence LockerReports, JSON, manifests, PDFs where present

ProvenanceArtifact source and lineage chains

ReplayEvidence-stage replay timeline

Manifest LineageSHA manifest references and linked files

Review Priority & Third-Party RTL Exercise

The Command Center features a dedicated Review Priority page representing a product-safe workflow layer. It displays prioritized target signals ranked by PageRank centrality, cone sizes, and feedback loop (Strongly Connected Component) involvement.

Third-Party RTL Exercise: Ranks targets on external public RTL modules (PicoRV32 and verilog-axi priority encoder).
Boundary: This benchmark does not claim vulnerabilities in PicoRV32 or verilog-axi. It demonstrates that the analyzer can process public third-party RTL and produce bounded review-priority rankings.

What Is Proven

A local dashboard exists and runs from the repository with npm start.
Evidence index lineage is summarized for partner review.
The dashboard reads data/dashboard_index.json instead of scraping frozen evidence artifacts at runtime.
Evidence summary, RTL Review Witness, affected cone summary, provenance, replay summary, manifest lineage, and limitations views are included.
The index includes references to existing evidence packets, Markdown reports, JSON reports, SHA manifests, screenshots, and known limitations where present.
The RTL Review Witness page now renders the public external RTL corpus exercise: PicoRV32 and verilog-axi controlled mutations, statuses, suspicious signals, cone counts, source paths, and result artifacts.
Screenshot export exists through npm run screenshots and produces a reproducible screenshot pack under assets/screenshots/.
A 5-minute demo script, evaluation kit README, known limitations document, and manifest are present.
The dashboard separates proof generation from proof visualization; it does not modify frozen reports, proof logs, synthesis outputs, manifests, or RTL.

What Is Not Proven

The dashboard is not a proof engine.
The dashboard is not a formal equivalence checker.
The dashboard is not a replacement for vendor EDA tools or human verification review.
The dashboard does not claim final verification closure, security signoff, whole-chip correctness, deployment readiness, or guaranteed Trojan detection.
The current evidence corpus is ARK-generated and public-corpus exercised where noted; customer-owned evidence ingestion has not yet been performed.
The current UI is a local demo/evaluation surface, not a hardened multi-tenant SaaS product.

Current Maturity

Local dashboard and evaluation kit exist; screenshot pack, generated evidence index, provenance view, replay view, manifest lineage, and evidence locker are implemented. The next maturity step is evaluator-owned evidence import: a partner supplies a small sanitized verification artifact set, and ARK maps it into the Command Center index for review.

A minimal evaluator artifact set consists of one or more Markdown or JSON verification reports, a manifest file, and optionally formal proof logs, cone artifacts, synthesis outputs, or waveform/trace summaries. ARK maps these into the dashboard index and delivers a configured local review packet.

What The Evaluator Gets On Day One

Dashboard sourceindex.html, src/app.js, src/styles.css

Index generatorscripts/build_dashboard_index.py

Generated indexdata/dashboard_index.json

Screenshot packassets/screenshots/

Evaluation docsREADME, demo script, limitations, manifest

Run commandnpm start from services/ark_chip/verification_dashboard

What The Evaluator Does Not Get Automatically

Full private ARK RTL, unrelated chip-IP source, unrestricted redistribution rights, hosted deployment rights, or customer-specific integrations are not included automatically. Deeper access is governed by paid evaluation terms, NDA scope, and field-of-use constraints.

Next Step

Paid evaluation: the evaluator supplies a small sanitized evidence folder containing reports, logs, manifests, proof outputs, or cone artifacts; ARK maps it into the Command Center index, delivers a local dashboard review packet, and reports whether the dashboard improves reviewability, replay lineage, and evidence navigation.