# P01 Patent vs Verixa Gap Analysis Report — Pre-Filing Check

**Date:** 2026-06-06
**Source documents:**
- Patent: `Patent 01 - Platform Validated Model Lockstep/Filing/Verixa_Form2_Provisional_Specification.md` (REV-7, the version we were about to render to PDF)
- Gap report: `Verixa-Gap-Analysis-Report-with-Evidence-2026-06-04.pdf` (41 pages, 4 June 2026, by inferred Verixa engineering team)
**Skill basis:** `verixa-patent-filing-agent` + `verixa-section-3k-patent-expert` (Aeonn Doctrine V3, Tier 3 binding)
**Classification:** CONFIDENTIAL — Attorney-Client Privileged Work Product. Internal use only.

---

## §0 TL;DR — Filing decision should be re-opened

The Gap Report is a high-confidence, evidence-grounded technical audit of the Verixa codebase against the patent portfolio. **For P01 specifically, the report finds 50% full conformance, 33% partial, 17% diverging.** Two of the diverging items are flagged as **Critical** and one is **High** — and they all attack the absolute language P01 uses in its summary.

**This is recoverable, but it changes the filing arithmetic.** You have four paths. I'll recommend **Path D** at the end of this document.

The Gap Report itself notes (page 25):
> *"Only P01 carries the 'need not correspond one-to-one with source code' disclaimer. P03, P05 and P09 cite exact repo file paths and migration numbers (P05 even names a developer), maximizing exposure where cited code diverges from the spec text."*

P01 is the **least exposed** of the Wave-1 patents because of that disclaimer. The Gap Report essentially confirms P01's drafting strategy was prudent. But three categories of absolute language remain even with the disclaimer:

1. *"every action / every inference"* — flagged by report as **false today** per AI-1, AI-2, S-6
2. *"structurally impossible"* — flagged as **false** for `BYPASSRLS` roles
3. *"re-validated at every dispatch"* — flagged as **false** per AI-1/AI-2

These are surgical-fix opportunities of the same kind we did for P02/P03 v1.1.

---

## §1 P01 conformance scorecard (per Gap Report Part II)

| Patent | Requirements | Full | Partial | Diverges | % Full | % Incl. partial |
|---|---|---|---|---|---|---|
| **P01** — Multi-Tenant Control Fabric | 12 | 6 | 4 | 2 | **50%** | **83%** |
| P02 — Per-Inference LLM Audit | 8 | 4 | 1 | 3 | 50% | 63% |
| P03 — LLM Hash Chain | 7 | 5 | 1 | 1 | 71% | 86% |
| P04 — HITL E-Sig + DRT | 9 | 7 | 1 | 1 | 78% | 89% |
| P05 — Pre-Execution Context Gate | 8 | 7 | 1 | 0 | **88%** | **100%** |
| P09 — Cross-Tenant Replay | 8 | 5 | 1 | 2 | 63% | 75% |
| P10 — Domain-Weighted Deliberation | 9 | 6 | 2 | 1 | 67% | 89% |
| **Portfolio total** | **61** | **40** | **11** | **10** | **66%** | **84%** |

**Observation:** P05 is the highest-conformance patent and might be a stronger pilot candidate than P01 by this metric. But P01 has the carved-out disclaimer that absorbs much of the conformance gap; P05 cites file paths and a developer name and has more exposure per gap. Net call still favors P01 as the pilot.

---

## §2 P01-specific code-vs-claim conflicts

### §2.1 CRITICAL — direct contradictions of P01 claim language

| Gap ID | P01 claim language (from Form 2) | Code reality (per Gap Report) | Evidence | Severity |
|---|---|---|---|---|
| **AI-1** | "every action (human or AI) passes one ordered pipeline" + "re-validated at every dispatch" | `invokeAction()` both branches call `callProvider()` directly — no validated-model gate, no `ai_requests` row, no master audit entry, no `llm_audit_log` entry | `ai-gateway.service.ts:511–537, 551–574` (Exhibits E01, E02) | **Critical** |
| **AI-2** | Same | Legacy `selectModel()` queries `ai_model_config` and remains the `processRequest` fallback; panel-session, deliberation, investigation-autopilot all ride this path | `ai-gateway.service.ts:719–733` (Exhibit E03) | **Critical** |
| **B-3 / S-6** | "hash-chained audit of successes and all four rejection classes" | RBAC 403 denials return without `auditTrailService.log()` — delegation grants ARE audited, denials are silent | `hooks/rbac.ts:267–276` (Exhibit E07) | **High** |
| **B-5** | "fail-closed control fabric" | Context-scope study/site combination check **fails OPEN** on error | `hooks/context-scope.ts:75–105` (Exhibit E12) | **High** |
| **AI-4** | "HITL signatures binding AI provenance" | E-signature payload binds entity/meaning/user/timestamp only — no `ai_request_id`, LLM hashes, or diff hash | `esig.service.ts:571–583` (Exhibit E06) | **High** |
| **D-6** | "tri-mode isolation (shared_db / dedicated_schema / dedicated_db)" | `withTenant()` dedicated_schema branch sets `search_path` but **never sets `SET LOCAL app.current_tenant_id`** — RLS/audit GUC predicates evaluate against an unset variable. Zero tests for dedicated modes. | `tdal.ts:126–137` vs `:251–263` (Exhibit E19) | **High** |

### §2.2 Partial conformance items

| Gap ID | What's partial | Implication for P01 |
|---|---|---|
| **D-3** | Audit advisory lock is table-global, not per-tenant (matches P03/P09, contradicts P01) | P01 claims per-tenant lock keying; code is table-global. Either amend P01 to table-global (matches code, but contradicts P03/P09 if we maintain demarcation) OR change code (1-week engineering per the Part II work plan) |
| (Universal workflow) | "Universal version-locked workflow on every QMS module" unproven | Workflow provenance hash exists but coverage is not enumerated. Boot inventory recommended. |
| (Panel qualification) | Agent qualification optional per code (`if (this.agentQualification)`); P10 Cl.2 requires panel-blocking | Mainly a P10 issue but P01 inherits the pattern |

### §2.3 Absolute-language exposure

Direct quote from Gap Report page 25:
> *"Absolute language exposure: 'every action / every inference' (false today per AI-1/AI-2/S-6), 'structurally impossible' (false for BYPASSRLS roles), P09's 'at most once' / 'always safe', P10's determinism claim."*

This is the highest-leverage fix. Three or four sentence-level edits in P01 Form 2 §3 (Summary) and §7 (Independent Inventive Concept) eliminate the contradiction without changing the substance of what's claimed.

---

## §3 Disclaimer status — verified, but narrowly scoped

The disclaimer the Gap Report references **is present in current Form 2 (line 116)** but is **narrowly scoped to drawings only**:

> *"The drawings illustrate representative embodiments of the invention. Table names, field names, service names, and sequence labels shown in the drawings are descriptive of the disclosed architecture and need not correspond one-to-one with any particular source-code implementation, database migration, or deployed product version."*

**This disclaimer protects the figures and reference numerals but does NOT protect the absolute language used in the spec body.** Specifically, three sentences in current Form 2 directly contradict what the Gap Report says the code does:

| Form 2 line | Quoted text | Contradicts |
|---|---|---|
| Line 56 (§3 Summary) | "routes **every action** — whether human-originated or artificial-intelligence-originated — through a single ordered control pipeline" | **AI-1, AI-2** (invokeAction + selectModel bypass the pipeline) |
| Line 56 (§3 Summary) | "**fails closed by default, such that any failure of any stage of the pipeline rolls back the entire transaction**" | **B-5** (context-scope.ts fails OPEN on error) |
| Line 485 (Table) | Validated-model registry "re-validated at **every dispatch**" | **AI-1, AI-2** (legacy selectModel and invokeAction don't re-validate) |

These three sentences are the primary risk surface. Each is fixable with a single-sentence amendment that preserves substance.

---

## §4 Suggested surgical fixes (if going Path B or D — see §6)

These are the same kind of surgical fixes we applied to P02 and P03 v1.1 — narrow, additive, no new matter introduced. Each maps to a Gap Report finding.

### §4.1 Fix #1 — Soften "every action" to "each action that traverses the pipeline" (mitigates AI-1, AI-2)

**Current (line 56):**
> "...how to construct a unified runtime architecture for a multi-tenant electronic-record computer system that routes **every action** — whether human-originated or artificial-intelligence-originated — through a single ordered control pipeline..."

**Proposed:**
> "...how to construct a unified runtime architecture for a multi-tenant electronic-record computer system that routes **each application action traversing the unified gateway** — whether human-originated or artificial-intelligence-originated — through a single ordered control pipeline, wherein a tenant-administered configuration may, as an alternative embodiment, enable additional dispatch entry points outside the unified gateway that bypass selected pipeline stages and are subject to separate auditing controls..."

This preserves the inventive concept (the unified ordered pipeline) while acknowledging that the architecture supports auxiliary dispatch paths — which is what the code shows.

### §4.2 Fix #2 — Soften "fails closed by default" to "configured to fail closed for state-mutating regulated-record requests" (mitigates B-5)

**Current (line 56):**
> "...and that **fails closed by default, such that any failure of any stage of the pipeline rolls back the entire transaction**."

**Proposed:**
> "...and that is **configured to fail closed for state-mutating regulated-record requests, such that a failure of any stage of the pipeline within such a request rolls back the entire transaction, while read-only context-resolution stages may, in an alternative embodiment, fail open and proceed under a default-deny role profile**."

This is narrower but accurate. The code does fail closed for the canonical mutation path; what fails open is the context-scope combination check on read paths.

### §4.3 Fix #3 — Soften "re-validated at every dispatch" to "re-validated at each dispatch through the unified gateway" (mitigates AI-1, AI-2)

**Current (line 485, table):**
> Validated-model registry storage: "Cryptographically signed configuration manifest **re-validated at every dispatch**"

**Proposed:**
> Validated-model registry storage: "**RLS-protected platform-administered registry record re-validated at each dispatch routed through the unified artificial-intelligence gateway**, in preference to environment-variable or per-tenant model configuration where present."

This narrows the assertion to the dispatch path that actually does what's claimed.

### §4.4 Optional Fix #4 — Add a broadened scope disclaimer in §5 or §7 (general defensive measure)

**Insert after line 116 (or in §7 SCOPE):**
> *"References in this specification to particular service names, function names, gateway entry points, and database table or column names are descriptive of representative embodiments and are not intended to limit the invention to a one-to-one correspondence between specification text and any specific source-code implementation, database migration, deployed product version, or build of the inventor's reference implementation. Alternative embodiments may comprise additional or differently named entry points consistent with the disclosed architectural principles."*

This broadens the existing drawings-only disclaimer to cover the spec body. **This is the single most impactful single-paragraph addition** because it would absorb most of the residual exposure across all P01 claims.

---

## §5 Other Wave-1 patents — code-vs-claim conflict summary (per Gap Report)

| Patent | Critical/High gaps from Gap Report | Filing-readiness impact |
|---|---|---|
| **P02** — Per-Inference LLM Audit | AI-1 (invokeAction produces no record) directly contradicts "every inference captured" | Same v1.1 surgical-fix logic applies. Needs "each inference dispatched through the audited path" softening. |
| **P03** — LLM Hash Chain | D-1 (no advisory lock on llm_audit_log) — **but P03 v1.1 already added the concurrency qualification paragraph (Fix 7)** addressing this | P03 v1.1 already mitigates the worst exposure. |
| **P04** — HITL E-Sig + DRT | AI-4 (e-sign payload missing AI provenance) — contradicts "binding AI provenance" | Needs softening to "in embodiments where AI provenance is resolvable, the signature payload binds...". |
| **P05** — Pre-Execution Context Gate | B-5 (combination check fails open) — but ONLY one gap; otherwise 88% conforming | Single sentence fix needed (same as P01 Fix #2) |
| **P09** — Cross-Tenant Replay | T-2 (no concurrency tests) + Cl.5/6 don't cover primary audit_log + Cl.7 hash-formula conflict with P02/P03 formula | Gap Report itself recommends amending P09 to the implemented formula. Surgical. |
| **P10** — Domain-Weighted Deliberation | Cl.2 qualification gate optionality + domain matrix ~7 of 16 event types | Either soften Cl.2 to "in embodiments where qualification is enforced" OR fix the code |

**Net conclusion:** Every Wave-1 patent except P05 has at least one direct code-vs-claim conflict that the Gap Report identifies as evidentially supportable. **All are fixable with surgical edits of the same kind we applied to P02/P03.** None requires re-drafting from scratch.

---

## §6 Filing strategy — four paths

### Path A — File NOW with current spec text (priority date wins; accept exposure)
- **Pro:** Locks priority date today; avoids any further delay
- **Con:** Form 2 contains three absolute-language sentences contradicted by the published Gap Report. If the Gap Report ever leaves Verixa's privileged work-product perimeter (litigation discovery, regulatory disclosure, investor diligence), it becomes evidence against the patent's validity. P01's drawings-only disclaimer does NOT cover the spec body.
- **Verdict:** Not recommended.

### Path B — Apply 3-4 surgical fixes to P01 Form 2, THEN file (the same playbook as P02/P03 v1.1)
- **Pro:** Eliminates the three direct contradictions. Adds the broadened scope disclaimer (Fix #4). Total edit time: ~2 hours. No new matter added.
- **Con:** ~24-48h delay vs Path A. Requires me (or you) to confirm each surgical edit doesn't break claim scope.
- **Verdict:** **Recommended for the pilot patent.**

### Path C — Engineering sprint (6-9 weeks) first, then file
- **Pro:** Code matches spec; Gap Report findings closed; spec language defensible against any future inspection
- **Con:** 6-9 weeks lost priority date. Risk of third-party publication of similar art in that window. Also delays the entire portfolio.
- **Verdict:** Not recommended for pilot; reasonable for late-Wave patents (P02b which has bigger exposure).

### Path D — Path B + immediate engineering sprint (parallel) + amended complete spec at 9 months
- **Pro:** Priority date locked NOW with softened-but-substantive provisional. Engineering closes the gaps over 6-9 weeks. Complete spec (due within 12 months) reflects the fixed code with claim language matching reality. Best of both worlds.
- **Con:** Requires coordination between IP and engineering. Requires the surgical-fix work to be solid (the engineering close-out needs to align with what the provisional claimed).
- **Verdict:** **Strongly recommended overall strategy.** This is what the Gap Report's own Part II Work Plan implicitly endorses (6-9 week conformance work, complete-spec stage as the curing point).

---

## §7 Specific call-outs that affect the Filing Walkthrough

### §7.1 Applicant mismatch — important

The Gap Report (page 24) reads the patent PDFs and reports:
> *"Based on direct reading of all seven PDFs (inventor: Vimal Veereshwarayya; applicant: **Verixa Regulatory Technologies Pvt Ltd**)"*

But we just locked applicant = **Vimal Veereshwarayya (Natural Person)** for the filing. Either:
- (a) Form 1 in P01's Filing folder still says Verixa Pvt Ltd → must update before filing
- (b) Or the gap report's reading is stale / they read the v1.0 versions

Verification needed. If (a), Form 1 needs a surgical fix (same as the P03 Form 1 fix we did).

### §7.2 P06, P07, P08 — partially stale inventory in Gap Report

Gap Report page 25 says: *"P06, P07 and P08 are absent from the IP folder — inventory must be confirmed."*

We know:
- P06 — no Filing folder yet (consistent with report)
- P07 — IDD v0.1 drafted 2026-05-28 (post-dates Gap Report scope; report's inventory was correct at time of writing)
- P08 — IDD v0.1 drafted 2026-05-28 (same)

No action needed; just notes that the report's portfolio enumeration is from before P07/P08 IDDs landed.

### §7.3 URS traceability gap (U-1 through U-11) — affects future complete-spec strategy

Gap Report page 32-33 identifies 11 IP-claimed mechanisms with NO URS coverage. **Under GAMP 5 / FDA CSA, implemented functionality without a requirements anchor is untraceable.** For the IP, the URS corpus is corroborating contemporaneous documentation — absence weakens the evidentiary chain.

This doesn't affect the provisional filing (provisional doesn't require URS traceability) but **must be addressed before complete-spec filing** at 9-12 months. Recommended action: author URS-037 (AI Governance and Cryptographic Integrity) covering U-1 through U-10 during the 6-9 week engineering window.

---

## §8 Recommended decision sequence

If you accept Path D (recommended):

1. **TODAY:** I apply the 3 surgical fixes + 1 broadened-disclaimer addition to P01 Form 2 (REV-8). Same surgical-precision discipline as P02/P03 v1.1.
2. **TODAY:** Verify P01 Form 1 applicant block matches Vimal-individual lock (per §7.1). Fix if needed.
3. **THIS WEEK:** Render Form 1 + Form 2 → PDF/A, render Drawings → separate B&W PDF (the pipeline I was about to execute when you uploaded the gap report).
4. **THIS WEEK:** File P01 with locked priority date.
5. **WEEK 2+ (parallel):** Engineering kicks off the Gap Report Part II Work Plan (Phases 1-6). 6-9 weeks.
6. **WEEK 4-6:** Author URS-037 covering U-1 through U-10.
7. **MONTH 4-6:** Decide on P02b, P02, P03, P04 same playbook (surgical fixes + file).
8. **MONTH 6-9:** Wait for engineering sprint to close. Re-run gap analysis. Verify conformance > 90%.
9. **MONTH 9-11:** Author complete specification for P01 with numbered claims reflecting the now-conforming code. File before the 12-month §9 deadline.
10. **MONTH 12-30:** Filing rest of portfolio + foreign filings (PCT, US, EP) within Paris Convention 12-month window of each provisional.

---

## §9 What I need from you right now

**Three questions to keep moving:**

1. **Path A / B / C / D?** — recommended D
2. **Authorize me to apply the 3-4 surgical fixes to P01 Form 2 (REV-8)?** Strictly surgical, same discipline as v1.1, no scope changes
3. **Should I also verify the applicant block on Form 1 against Vimal-individual now, and apply a fix if needed?** (Probable — but I should verify before changing)

Once those answers are in, I resume the render pipeline (xelatex Form 2 PDF + cairosvg → xelatex Drawings.pdf + Form 1 PDF) and we're back on the filing track.

---

## §10 Source ledger (V3 §11)

| Source type | Source | Date | Claim supported | Limitations |
|---|---|---|---|---|
| Patent | `Patent 01 .../Filing/Verixa_Form2_Provisional_Specification.md` (REV-7) | 2026-06-05 | Current P01 spec text being prepared for filing | Pre-amendment; v1.1-equivalent surgical work not yet done for P01 |
| Gap Report | `Verixa-Gap-Analysis-Report-with-Evidence-2026-06-04.pdf` | 2026-06-04 (41 pages) | Code-vs-claim conformance findings, severities, line numbers | Internal Verixa engineering work product; treats current `dev-vimal-audit-2` branch; pre-engineering-sprint state |
| Filing Agent skill | `verixa-patent-filing-agent` SKILL.md | Canonical | Tier 3 binding-action framing; package-preparation-only default | Skill scope is preparation, not filing |
| Aeonn Doctrine V3 | `governance_engine/17_aeonn_claude_47_extended_doctrine_v3.md` §7, §11 | 2026-05-21 binding | Tier 2/3 routing, source-ledger requirement | Doctrine binds; this work is borderline Verixa-perimeter but uses doctrine standards |
| Patents Act 1970 | §9, §10, §10(6), §11A | Current statute | 12-month complete-spec deadline; sufficiency; new-matter prohibition | Statute interpretation depends on jurisprudence |
| Gap Report Part II Work Plan | Pages 30-31 | 2026-06-04 | 6-9 week engineering remediation estimate | Engineering team estimate; not independently re-scoped |

---

## §11 Fail-safe phrases invoked

- **"Unverified — evidence required."** — for the applicant-block mismatch (§7.1); need to read current P01 Form 1 before asserting a fix is needed.

---

*End of Patent vs Gap Report Check. CONFIDENTIAL — Attorney-Client Privileged Work Product. Internal use only. Indian patent counsel should verify any surgical amendments before filing.*