The Claims Register
Every claim we make in public has a row here: what it says, where the evidence lives, what its status is today, and the specific gate it must pass before we may describe it as current.
Rows are not marked current because we believe them. They are marked current when they pass their gate.
Full evidence and amendment history ↓
Evidence baseline: 22 July 2026 · Production amendments: 3–14 August 2026. The register preserves historical status rather than rewriting it: rows 1–12 retain their original 22 July evidence state; rows 13–33 record production capabilities and runtime evidence verified between 3 and 14 August 2026 — rows 13–15 (CP-001, CP-002, the Claude ↔ Gemini provider swap) on 3 August, row 16 (CP-003) on 5 August, row 17 (Pitch Deck) and rows 18–19 (Resume, Cover Letter) on 6 August, rows 20–21 (Executive Summary, Landing Page) on 7 August, row 22 (the eight-capability aggregate) on 8 August, row 23 (Software Test Cases) and row 24 (the nine-capability aggregate, superseding row 22's count without editing it) on 9 August, row 25 (Pricing Strategy) and row 26 (the ten-capability aggregate) on 10 August, rows 27–28 (Landing Page Audit, Professional Email) and row 29 (the twelve-capability aggregate, superseding row 26's count without editing it) on 11 August, rows 30–32 (Project Management, the thirteen-capability aggregate superseding row 29, and WAR-CROSS-001's cross-provider execution result) on 13 August, and row 33 (the runtime's own provider abstraction, distinct from WAR-SPEC-003's unmet compiler gate) on 14 August — against the WAR Launch Narrative's own freeze. The production inventory is revalidated against current launch evidence before external publication. When a row's status changes, the row changes — the register is not rewritten.
- Implemented is described as current.
- Not yet implemented is described explicitly as specified, in development, or roadmap.
- We never imply an application runs on WAR unless that is technically true today.
- Enterprise Context Intelligence is described as the first enterprise application of the WAR architecture and its commercial validation pathway — that wording only.
| # | Claim | Evidence source | Status today | Gate to “current” |
|---|---|---|---|---|
| 1 | Intent Graph: typed, event-sourced, hash-verifiable, language-free | WAR-SPEC-001 (RFC-2119, implementable without founders) |
SPECIFIED
Specified; pre-prototype |
Prototype passes ir-lint invariants 1–8 |
| 2 | Intent Understanding Layer: 7-stage pipeline; the only component seeing language | WAR-SPEC-002 |
SPECIFIED
Specified; pre-prototype |
S1–S7 running on real utterances |
| 3 | Deterministic planner — specified as a pure function; byte-identical replay not yet publicly claimed | WAR-SPEC-003, Rule PL-1 |
SPECIFIED
Specified; flagged as the spec's riskiest promise |
CI replay test green |
| 4 | Provider compilers; cross-model replay; mid-execution fallback | WAR-SPEC-003 §5, WAR-TRACE-001 T12 |
SPECIFIED
Specified; worked on paper |
Two compilers passing identical artifact contracts |
| 5 | Clarification Engine (taxonomy, budget, why-carrying questions) | WAR-SPEC-004 Part A |
SPECIFIED
Specified; pre-prototype |
Live clarification round-trip |
| 6 | Build Along™: blast radius, interruption, rendered narration, replay | WAR-SPEC-004 Part B, WAR-TRACE-001 T10–T12 |
SPECIFIED
Specified; worked on paper |
Live interrupted session |
| 7 | Hinglish kirana end-to-end trace | WAR-TRACE-001 |
SPECIFIED
Complete worked trace on paper — not an execution log |
Prototype reproduces the trace |
| 8 | Pivot-language verification harness | WAR-SPEC-002 §5 |
SPECIFIED
Designed; not in CI |
Harness running on native-authored corpus |
| 9 | WAR development process includes machine-checked doctrine gates and self-critiquing specifications | North Star; DOS-user doctrine; WAR-TRACE-001 Coda F1–F5 (the trace found 5 defects in its own specs) |
LIVE
Current — demonstrable today |
— |
| 10 | Enterprise Context Intelligence executes on WAR | — |
NOT CLAIMABLE
NOT claimable. ECI is the first enterprise application of the WAR architecture and its commercial validation pathway. It does not currently execute through the WAR runtime. |
ECI executing via the WAR pipeline |
| 11 | WAR public platform launch | Founder target |
IN DEVELOPMENT
In development; launch target 15 August 2026. Production capabilities already running before public launch — see rows 13–24. |
Launch |
| 12 | Cross-provider equivalence demonstration | — |
PLANNED
Committed milestone for Finals (27–29 October 2026) |
Demo rehearsed on a real build |
| 13 |
CP-001 Business Website executes on the WAR runtime Added 3 Aug 2026
|
CP-001 freeze declaration, 3 August 2026 |
LIVE
Live |
— |
| 14 |
CP-002 Company Research Brief executes on the WAR runtime Added 3 Aug 2026
|
Provider Runtime v1, 3 August 2026 |
LIVE
Live |
— |
| 15 |
Claude ↔ Gemini provider swap by configuration, with no capability-specific code Added 3 Aug 2026
|
Provider Runtime v1, 3 August 2026 |
LIVE
Live |
— |
| 16 |
CP-003 Business Proposal executes on the WAR runtime Added 5 Aug 2026
|
CP-003 Document Studio v1.2 freeze, 5 August 2026 |
LIVE
Live |
— |
| 17 |
Pitch Deck executes on the WAR runtime Added 6 Aug 2026
|
CP-004 production click-through, 6 August 2026 |
LIVE
Live |
— |
| 18 |
Resume executes on the WAR runtime Added 6 Aug 2026
|
Real production click-through, 6 August 2026 |
LIVE
Live |
— |
| 19 |
Cover Letter executes on the WAR runtime Added 6 Aug 2026
|
Real production click-through, 6 August 2026 |
LIVE
Live |
— |
| 20 |
Executive Summary executes on the WAR runtime Added 7 Aug 2026
|
Real production click-through, 7 August 2026 |
LIVE
Live |
— |
| 21 |
Landing Page executes on the WAR runtime Added 7 Aug 2026
|
Real production click-through, 7 August 2026 |
LIVE
Live |
— |
| 22 |
Eight capabilities, across four of WAR's five Studios, execute on one runtime Added 8 Aug 2026
|
WAR Launch Narrative v1.0 §14 (“What Is Live”), frozen 8 August 2026 |
LIVE
Live — none of the eight required rebuilding the runtime |
— |
| 23 |
Software Test Cases executes on the WAR runtime Added 9 Aug 2026
|
Real production evidence: two real click-throughs against a real confidential technical document (an SDK integration guide), the second of which surfaced a credential-leak, an unsupported-inference, and an unsupported-requirement finding — all found, fixed, and re-verified against the same evidence before this row was added, 9 August 2026 |
LIVE
Live |
— |
| 24 |
Nine capabilities, across four of WAR's five Studios, execute on one runtime Added 9 Aug 2026
|
Supersedes row 22's count, not its evidence — row 22 is preserved unedited. 9 August 2026 |
LIVE
Live — none of the nine required rebuilding the runtime |
— |
| 25 |
Pricing Strategy executes on the WAR runtime Added 10 Aug 2026
|
Real production evidence: two real click-throughs against a real scenario (an AI video editing tool with a stated cost floor and competitor range), the first of which surfaced a real defect — an over-conservative no-fabrication rule caused the model to refuse any concrete price recommendation — fixed and re-verified on the second click-through (real tier prices with the margin math shown). A separate shared-renderer defect (a numbered list losing its own numbering around nested sub-bullets) was found and fixed along the way, 10 August 2026 |
LIVE
Live |
— |
| 26 |
Ten capabilities, across five of WAR's five Studios, execute on one runtime Added 10 Aug 2026
|
Supersedes row 24's count, not its evidence — row 24 is preserved unedited. Pricing Strategy is Strategy Studio's first Live capability, the first time all five Studios are represented. 10 August 2026 |
LIVE
Live — none of the ten required rebuilding the runtime |
— |
| 27 |
Landing Page Audit executes on the WAR runtime Added 11 Aug 2026
|
Real production evidence: a real click-through against a real external page, verified against an independent reference audit of the same page — eight production defects found only by running it for real and reading the artifact, all found, fixed and re-verified before this row was added, 11 August 2026 |
LIVE
Live |
— |
| 28 |
Professional Email executes on the WAR runtime Added 11 Aug 2026
|
Real production evidence: three real click-throughs, six defects found and fixed across them — none caught by the pack's own passing test suite, 11 August 2026 |
LIVE
Live |
— |
| 29 |
Twelve capabilities, across five of WAR's five Studios, execute on one runtime Added 11 Aug 2026
|
Supersedes row 26's count, not its evidence — row 26 is preserved unedited. Landing Page Audit and Professional Email, both added 11 August 2026. |
LIVE
Live — none of the twelve required rebuilding the runtime |
— |
| 30 |
Project Management executes on the WAR runtime Added 13 Aug 2026
|
Real production evidence: a real click-through against a real domain ("Track my job applications") — the generated board's column structure was itself inferred from that description, not a fixed template, 13 August 2026 |
LIVE
Live |
— |
| 31 |
Thirteen capabilities, across five of WAR's five Studios, execute on one runtime Added 13 Aug 2026
|
Supersedes row 29's count, not its evidence — row 29 is preserved unedited. Project Management, added 13 August 2026. |
LIVE
Live — none of the thirteen required rebuilding the runtime |
— |
| 32 |
The same WAR Capability Pack can execute through multiple AI providers without capability-specific provider coupling Added 13 Aug 2026
|
WAR-CROSS-001 — Project Management, the same Capability Pack, executed live through three independent provider integrations (Gemini, Claude, OpenAI), each producing a real generated artifact, with provider-specific request/response differences handled entirely at the provider boundary and no change to the capability itself. 13 August 2026 |
LIVE
Live |
— |
| 33 |
WAR contains a working provider abstraction, used by production Capability Packs, independent of any single AI provider Added 14 Aug 2026
|
WAR-CROSS-001 — the same production Capability Pack (Project Management) routed through this runtime abstraction to execute live through Gemini, Claude and OpenAI, with no capability-specific provider implementation changes. Distinct from row 32 (a Capability Pack's own portability) and from WAR-SPEC-003's deeper Intent-Graph compiler architecture (rows 3-4), which retains its own independent, unmet gate. Evidence dated 13 August 2026. |
LIVE
Live |
— |
Not yet multilingual
Pivot-language verification is designed but not yet in CI. No multilingual determinism demonstration is published yet because the verification harness has not cleared its CI gate.
Not yet an enterprise runtime
ECI is the first enterprise application of the architecture but does not currently execute through WAR. This is why there is no Enterprise section in the navigation: a nav item with only promises behind it is worse than no nav item.
Currently demonstrable
WAR is running in production through real Capability Packs. Users can submit natural-language goals, receive inferred workflows and deterministic blueprints, and generate working artifacts. WAR-CROSS-001 has additionally demonstrated the same Project Management Capability Pack through Gemini, Claude and OpenAI.
The research discipline itself is also demonstrable: WAR-TRACE-001, a complete end-to-end worked trace on paper, surfaced five defects in its own specifications when run against them.
A system with the property does not ask to be believed. It offers to be checked.
Every status chip elsewhere on this site links back to a row on this page. If a public claim cannot be traced to a row here, treat it as a defect in the site.