> Neurobrick is an AI governance architecture that separates AI decision proposals from execution authority. Implemented in multi-asset financial research and execution environments. Inspect dated engineering results, source and test populations, alongside an unchanged archive of original system and technical records. [neurobrick]() [principles]() [architecture]() [capabilities]() [authority]() [observation]() [evidence]() [engineering]() [verification]() [scope]() Technical evidence and fixed records # Evidence you can inspect. History you can identify. Public engineering results retain their source, measured population and limits. Original records stay available at their original dates; a new summary does not rewrite an earlier observation. Scoped engineering review ## Implemented paths and dated controls Control states, specific refusals and source implementations are summarized below with their original dates, environments and integrity references. [Follow summary, scope and source →]() Verification method ## The observer must preserve uncertainty Output interpretation and failure cleanup are tested on their actual paths, including negative controls that reject broken implementations. [Original record →]() [SHA-256 →]() Finding history and later observations The 2026-10-10 reconciliation links preserved findings to later source and candidate CI observations. It does not establish operating closure. [Finding summaries and current status →]() [Original reconciliation JSON →]() [File integrity SHA-256 →]() Evidence index ## Summary → scope → original record. Read what was implemented or observed, then follow the record that supports the exact claim. Different windows, environments and test populations remain separate. ### Portfolio kill-switch state Implemented and observed in the stated scope Record observed/reviewed · 1 October 2026 The inspection reported the portfolio kill switch as ARMED, separately from the approval-axis summary. Environment / scope: Control-plane state report in the original inspection. Limit: ARMED is a reported configuration state. This record does not demonstrate a triggered portfolio stop, its reach across every order path, or the current state. [Read the original supporting JSON →]() [Verify the original file SHA-256 →]() **Evidence identity, source and integrity** **Evidence ID** neurobrick-system-review-2026-10-01 **Original JSON pointers** /observed\_states/kill\_switch **Source commits by recorded role** Not recorded in this source; no commit is inferred. **Original file SHA-256** 399267ec9e1d081ddc33d2a42f13e0144038718a62fc2fd469db621258bbe252 **Current operating coverage** Not verified by this presentation. A checksum identifies the file and does not independently reproduce its private run. ### Server-side order refusal Implemented and observed in the stated scope Record observed/reviewed · 27 September 2026 The original inspection records identified order attempts refused by the server-side risk check and the core sell block. Environment / scope: Identified primary-engine attempts in the record's stated period. Event window stated in the original: 26 September (KST). The record date above is the inspection date. Limit: This is an internal dated observation on a particular execution context. The executing commit was not recorded; accepted-case testing, universal enforcement and present operating state are not established by it. [Read the original supporting JSON →]() [Verify the original file SHA-256 →]() **Evidence identity, source and integrity** **Evidence ID** neurobrick-system-review-2026-09-27 **Original JSON pointers** /authority/3/detail **Source commits by recorded role** Not recorded in this source; no commit is inferred. **Original file SHA-256** b1021299dd28c133c3ffb121463722add5d6dbc3741c4a95f476a5ef7e35135b **Current operating coverage** Not verified by this presentation. A checksum identifies the file and does not independently reproduce its private run. ### Opt-in, lease and certified-trial conditions Implemented and observed in the stated scope Record observed/reviewed · 27 September 2026 Every autonomous submission attempt in the stated half-hour was skipped while opt-in, admission-lease and certified-trial conditions were inactive or not ready. Environment / scope: Autonomous submission attempts in the inspected half-hour. Limit: This records refused or skipped progression, not successful issuance, certification or release. Development/testnet observations elsewhere refer to another service and must not be read as this same path changing over time. [Read the original supporting JSON →]() [Verify the original file SHA-256 →]() **Evidence identity, source and integrity** **Evidence ID** neurobrick-system-review-2026-09-27 **Original JSON pointers** /capabilities/1/detail **Source commits by recorded role** Not recorded in this source; no commit is inferred. **Original file SHA-256** b1021299dd28c133c3ffb121463722add5d6dbc3741c4a95f476a5ef7e35135b **Current operating coverage** Not verified by this presentation. A checksum identifies the file and does not independently reproduce its private run. ### Kill and release event identity Implementation present; runtime effect not verified Record observed/reviewed · 30 September 2026 The source review records event identity in the kill and release producer. Environment / scope: Merged and installed source inspected in the original review. Limit: The inspected live ledger was empty and no event was triggered for that review. An actual event identifier and natural trigger/clear execution remain unknown. [Read the original supporting JSON →]() [Verify the original file SHA-256 →]() **Evidence identity, source and integrity** **Evidence ID** neurobrick-engineering-progress-2026-09-30 **Original JSON pointers** /earlier\_advances/1/detail **Source commits by recorded role** - This recorded claim: bc89dacf89680597bb1be8d833a79128be103aa8 · EmergencyKillSwitch.trigger / clear **Source symbol** EmergencyKillSwitch.trigger / clear **Original file SHA-256** f3f85024f0a1a8a03285aa27b836130c038d2b0d020057650b85cbf94d7725ee **Current operating coverage** Not verified by this presentation. A checksum identifies the file and does not independently reproduce its private run. ### Observer output and process cleanup Implementation present; runtime effect not verified Record observed/reviewed · 2 October 2026 The recorded selection passed 18 cases and rejected six defective controls. It separates actual child behavior from controlled cleanup dependencies. Environment / scope: Private isolated process-observer validation. Limit: The record does not prove descendant reaping, current installation or natural operating adoption. The core repository is private; the public checksum does not reproduce the private run. [Read the original supporting JSON →]() [Verify the original file SHA-256 →]() **Evidence identity, source and integrity** **Evidence ID** neurobrick-verification-2026-10-02 **Original JSON pointers** /verification/passed · /verification/skipped · /verification/negative\_controls\_rejected **Source commits by recorded role** - This recorded claim: d6e06a514f5bf07ad7d1d3205975c87a9d611cd9 · kill\_tree / run\_pytest\_once **Source symbol** kill\_tree / run\_pytest\_once **Original file SHA-256** f215c307fcdd340194db32b7b8791147ad8f2da1158ca91c17cb716230535ddd **Current operating coverage** Not verified by this presentation. A checksum identifies the file and does not independently reproduce its private run. ### Committed risk-verdict candidate Implementation present; runtime effect not verified Record observed/reviewed · 4 October 2026 The direct risk-verdict selection changed from 16 failures to 18 passing cases. The broader selection retained 431 existing cases and passed 433 after the patch. Environment / scope: Private isolated Guardian candidate and compatibility selection. Limit: The direct 18 cases are included in the broader 433, so they are not added together. Candidate CI is distinct from later-main compatibility, current consumer identity and natural operating effect. [Read the original supporting JSON →]() [Verify the original file SHA-256 →]() **Evidence identity, source and integrity** **Evidence ID** neurobrick-engineering-2026-10-04 **Original JSON pointers** /guardian/direct/before/failed · /guardian/direct/after/passed · /guardian/focused/after/passed · /guardian/focused/existing\_cases\_retained · /guardian/direct\_is\_subset\_of\_focused **Source commits by recorded role** - This recorded claim: e9a241aca95ffa8713fad5d56b2fc6280a0ffe75 **Original file SHA-256** 64f5a5b70d7c616cab850fd5b84c4696f027cd18cad589ae40e09fa0d9f7804d **Current operating coverage** Not verified by this presentation. A checksum identifies the file and does not independently reproduce its private run. ### Input measurement and decision identity Implementation present; runtime effect not verified Record observed/reviewed · 30 September 2026 The review covers input measurement, observation freshness and decision identity through their producer and consumer paths. Environment / scope: Frozen local source candidates with separately recorded populations. Limit: The measurement, freshness and lineage selections have different sources and may overlap. Their counts are not a whole-system total, and local validation is not natural operating adoption. [Read the original supporting JSON →]() [Verify the original file SHA-256 →]() **Evidence identity, source and integrity** **Evidence ID** neurobrick-engineering-progress-round2-2026-09-30 **Original JSON pointers** /improvements/0/tests · /improvements/1/tests · /improvements/2/tests **Source commits by recorded role** - Input measurement: f3cadaedbaf3a0c00acc34c3def18780d153169c · \_measure\_jsonl\_multi / \_assess\_input / run\_once / read\_research\_lab - Observation freshness: 1fdceaac6d4ab173b38f50ff5552457b72eac28a · build\_freshness / build\_receipt / render - Decision identity: 87400402b7839632bed2bd8fbefa4b0a2856cf82 · Node.\_\_post\_init\_\_ / Edge.\_\_post\_init\_\_ / own\_decision\_identity; R-K1 → R-K4 → BuildContext.absorb → metrics **Original file SHA-256** 3c1fe70623b60f3d325b1e7bc9359c18236bf0cb84726e98399af6c221f75d97 **Current operating coverage** Not verified by this presentation. A checksum identifies the file and does not independently reproduce its private run. ### Heartbeat, provenance and recovery observation Implementation present; runtime effect not verified Record observed/reviewed · 30 September 2026 The source selections cover heartbeat validity, retained observation time and direct affected-component recovery evidence. Environment / scope: Private focused producer-to-consumer validation. Limit: These are separately scoped internal selections. Historical publication runs also recorded skipped cases; those cases were not evaluated. No natural deployment or recovery result is created by presenting these tests. [Read the original supporting JSON →]() [Verify the original file SHA-256 →]() **Evidence identity, source and integrity** **Evidence ID** neurobrick-operational-proof-publication-2026-09-30 **Original JSON pointers** /improvements/0/independent\_test\_count · /improvements/1/independent\_test\_count · /improvements/2/independent\_test\_count **Source commits by recorded role** - Heartbeat validation: 9fa70597d98f91c850b549258bd5dbf9488caf45 · \_state\_age / run\_once\_sync / read\_daemon\_health / \_section\_daemons - Retained observation provenance: 5918e95232754e48d178f37249528526215c2dc5 · \_read\_capital\_flows / evaluate / read\_trading - Affected-target recovery observation: 3cc129f4a2675855843dc367ed8aadc5de3e20d0 · daemon\_health\_master.evaluate / \_recovery\_health\_in\_window / auto\_recovery\_effectiveness\_tracker.evaluate **Original file SHA-256** 57d41bba421bbe3fe853c78153746e183ba755e90745432de7b165133fbf2624 **Current operating coverage** Not verified by this presentation. A checksum identifies the file and does not independently reproduce its private run. ### Stable identity for evaluation and selection receipts IMPLEMENTED NOT RUNTIME VERIFIED **What changed** Entry-code fingerprints record an explicit serialization format. Selection replay refuses unknown formats and retains the historical comparison path. **Why it matters** Keeps evaluation and selection observations tied to the entry code that produced them while preserving uncertainty. **Implemented** Pinned evaluation-input and forward-selection receipt modules. Tracker integration and natural runtime adoption are outside this local test population. **Implementation status** IMPLEMENTED\_NOT\_RUNTIME\_VERIFIED **Observed** The supplied local test record contains 8 passed cases out of 8. Natural runtime adoption is not established. **Environment / scope** Isolated Python stdlib modules with synthetic pure callables Evaluation entry-code identity and selected replay counterexamples; excludes tracker integration and natural runtime **Evidence ID** evo-67b239088702e7c706f9 **Source commit** 4e9700fdfb8a9dcfc6049d74b0157437d19af3fc **Original observation date** 2026-10-09T11:47:53.594795+00:00 **Canonical record SHA-256** 67b239088702e7c706f9ddaaf755b260fb77e65074934f9cfd3db60491d61958 **Remaining limitations** No natural runtime or live adoption observed. Entry-code fingerprints do not attest imported dependencies or globals. Pre/post equality does not rule out change-and-restore races. Selection replay does not reproduce statistics, queue effects or execution authority. The pipeline validates recorded identity and consistency; producer authenticity and implementation meaning require human review. **Run artifact SHA-256** 6d0a3fbe23089d0750dbc334bacd39e798f2d0a7f26481655b1bcd71e93380e0 [Public source and test records →]() [Archive file SHA-256 →]() ### Unevaluated inputs no longer count as clean evidence IMPLEMENTED NOT RUNTIME VERIFIED **What changed** Freshness rates count only explicitly evaluated states. Soak readiness requires a recognized comparison-record kind instead of accepting any JSON object. **Why it matters** Prevents missing evaluation from being reported as a clean measurement or as evidence that a comparison ran. **Implemented** Receipt quality calculation, displayed measurement scope and comparison-observation recognition at the pinned source commit. Original input rows, schemas and readiness thresholds are preserved. **Implementation status** IMPLEMENTED\_NOT\_RUNTIME\_VERIFIED **Observed** Source inspection only; natural runtime is not established. **Environment / scope** Pinned source inspection of receipt quality and comparison-observation functions build\_receipt, render and soak\_maturity only; operating scan adoption is outside this record **Evidence ID** evo-4a2d40d3a7638c025c39 **Source commit** 8eb4eee4118d8d037db06a1941708e509d6aa0a0 **Original observation date** 2026-10-22T11:46:26.993Z **Canonical record SHA-256** 4a2d40d3a7638c025c3930c8a4a7504aeb1567f44a493cd9b1d47662e5c31e64 **Remaining limitations** No natural operating scan adoption or current collector freshness was verified. Recognizing a comparison kind does not authenticate its producer or payload. This correction does not restore older ledger records, establish live trading coverage or enable external anchoring. The pipeline validates recorded identity and consistency; producer authenticity and implementation meaning require human review. [Public source and test records →]() [Archive file SHA-256 →]() How to interpret status ## Not observed ≠ not implemented. A verification limitation qualifies a particular claim. It is not automatically a verdict that the underlying function does not exist. Source presence alone also does not prove a reachable or verified execution path. ### Implemented and observed in the stated scope IMPLEMENTED\_AND\_OBSERVED A specified behavior or state was observed in the recorded environment and window. This is not universal coverage or a fresh live-state claim. ### Implementation present; runtime effect not verified IMPLEMENTED\_NOT\_RUNTIME\_VERIFIED Source, an installed-source inspection or a local candidate exists. The relevant natural consumer behavior or current operating effect has not been established. ### Partial implementation or integration PARTIALLY\_IMPLEMENTED Named paths exist, while another named part is incomplete, sealed or not connected. Read the exact gap rather than treating the whole system as absent. ### Design only for this claim DESIGN\_ONLY An intended responsibility or interface is described, without an implementation bound to this particular claim. ### Unknown in the examined scope UNKNOWN The necessary source, observation or consumer identity was not available or not read. Unknown is neither permission nor evidence that a function is missing. ### Not observed is not not implemented Implementation existence, path reachability, a dated observation and verified operating effect are separate questions. Evidence that was not read cannot establish absence. ### Not implemented This requires evidence about a named searched code or execution scope. An unread artifact, a missing runtime observation or a design limitation alone is insufficient. ### Observed in a limited scope Keep the specific behavior, environment, original time and measured population. A refused attempt is a narrower observation than testing all accepted and refused cases. ### Verified for a specific execution path The actual consuming path, applicable inputs and accepted/refused or defective/corrected cases must be bound to the tested source and population. ### Not universally established Evidence for one path can coexist with missing coverage across other engines or markets. It does not erase the path that was actually examined. **Archived system inspection** Inspected operating footprint ## 7 compute hosts. 1 shared storage appliance. **Compute hosts** 7 Control, execution, inference, observation and CI **Storage** 1 NAS Shared storage; counted separately from compute **Execution systems** 3 Independent logical execution boundaries **Control services** 7 Running containers on one control host Mesh registration: 9 devices, 9 online. These are network identities, not the physical operating-box count. A dedicated-fabric compute host is outside this mesh; inactive client devices are inside it. Read-only inspection of the known operating fleet. Point-in-time internal disclosure, not live telemetry or independent attestation. [System review JSON]() [SHA-256]() [Historical evidence]() This record contains the original service, input-lineage, permission-setting and infrastructure observations. It is a fixed internal inspection, not live telemetry or an independent attestation. [Complete original record →]() [SHA-256 →]() Complete public evidence archive ## Original artifacts. Original checksums. All original files remain available. Their dates, populations, field meanings and disclosure limits are historical facts; they are not current website acceptance goals. Engineering · record 01 [Original supporting JSON]() [File integrity SHA-256]() Verification · record 02 [Original supporting JSON]() [File integrity SHA-256]() System review · record 03 [Original supporting JSON]() [File integrity SHA-256]() System review · record 04 [Original supporting JSON]() [File integrity SHA-256]() Operational proof · record 05 [Original supporting JSON]() [File integrity SHA-256]() Operational proof publication · record 06 [Original supporting JSON]() [File integrity SHA-256]() Engineering progress · record 07 [Original supporting JSON]() [File integrity SHA-256]() Engineering progress round2 · record 08 [Original supporting JSON]() [File integrity SHA-256]() System review · record 09 [Original supporting JSON]() [File integrity SHA-256]() System review · record 10 [Original supporting JSON]() [File integrity SHA-256]() Development review · record 11 [Original supporting JSON]() [File integrity SHA-256]() Current · record 12 [Original supporting JSON]() [File integrity SHA-256]() Pm receipt · record 13 [Original supporting JSON]() [File integrity SHA-256]() Am · record 14 [Original supporting JSON]() [File integrity SHA-256]() How to read the evidence A checksum identifies the public file. It does not reproduce a private run. Local tests establish the specific behavior and population recorded; deployment, natural consumer behavior and external reproduction require separate observations. [Public scope and technical limits →]() Neurobrick · systems, authority and evidence Last updated · 22 October 2026 [Public scope →]() [Canonical page]()