BLACKMORE / TECHNOLOGY PORTFOLIO

NIKI — AI Creation and Verification Council

NIKI is the research and invention intelligence programme in BTG’s Knowledge Capital Economy: it is oriented toward executable discoveries, not just conversational answers.

Blackmore Technology Group approved emblem

How NIKI could supply the data exchange

NIKI is BTG's computational invention intelligence: a research candidate can become more than an answer in a conversation. If its executable method passes independent technical and rights qualification, it may become a distinct data-derived intellectual asset with an issuer record and a rights-defined Bit Trade Global ticker. The Exchange is responsible for any permitted listing and trading; NIKI does not issue securities or establish market value.

How Bit Trade Global lists, values and trades eligible data assets →

Purpose and the research loop

NIKI is intended to choose or receive an authorized technical problem, use eligible source knowledge, produce candidate implementations, test the candidates against defined requirements and retain the outcome and its provenance. A claimed invention is meaningful only when it can execute, be evaluated against an independent baseline and survive repeatable checks.

A research system that can generate a plausible explanation has not necessarily generated a working algorithm. BTG’s approach emphasizes evidence about the executable artifact, test conditions and limitations rather than the persuasiveness of a text response.

Qualified outputs

The reusable outcome of a successful experiment may be a DataChip, a circuit arrangement, a revised genome, a dataset annotation or a bounded scientific finding. The output needs version identity, dependencies, testing evidence, allowed inputs and the conditions under which it is effective.

An unsuccessful experiment can still contribute valuable knowledge by showing which proposed arrangement failed and in what circumstances. Keeping that negative evidence reduces wasteful repetition and informs later research.

Permission and autonomy

NIKI’s research activities are subject to ownership, safety, resource and policy boundaries. An agent can propose tests or changes without automatically receiving permission to access protected information, publish a commercial asset or alter a production authority record. ADAM and underlying organizational controls separate recommendations from authorized execution.

Relationship to the research network

FarmNodes provide logical laboratories or bounded workflows; BTDU supports controlled knowledge reconstruction; DataChips make working abilities reusable; ENTITY supplies provenance and rights context; Verifiable Reality distinguishes each kind of proof. This architecture is designed to let specialized experiments inform the broader programme without erasing their origin.

Evidence and availability

BTG’s October 2026 source material discusses bounded development results. Those records do not establish general self-directed invention across all subject areas, independent laboratory qualification or a universally supported commercial API. Prospective partners can assess a narrowly defined experiment and its real test evidence before making operational commitments.

What NIKI testing demonstrates

A past narrow 65,536-input sensor execution matched its bounded oracle in local WASM receipts. A separate October 19 sensor-consensus candidate failed 57 of 65,536 static cases. In the distinct October 20 Granby task, the initial algorithm failed 256 of 65,536 inputs and the first two repairs returned unchanged source. Fourth and fifth model responses created changed source, but their records show source_executed=false and qualification=false. The objective is autonomous generation, repair, independent validation and retention of qualified executable abilities, but these observations do not demonstrate a universally self-improving production intelligence.

Explore the dated engineering evidence, public ledger and instrument records, and professional evaluation inquiries.

HISTORICAL RESEARCH EVIDENCE · 23 OCTOBER 2026 · 18:26 UTC

NIKI AI Council and FarmNode qualification

Earlier 18:26 UTC audit: 114/114 defined local tests passed across nine suites, including the earlier 70 tests. The four zero-priced hosted model entries were catalogue-checked; no authenticated hosted model comparison was proven. Three sector baseline packs processed 25 original historic data observations, not 25 new inventions. Security gate tests passed, but there are still no independently enrolled physical FarmNodes.

Original NIKI Granby code failed 256/65,536 cases. A distinct engineer-reviewed deterministic correction passed 65,536/65,536 in isolated WASM; it was not autonomous NIKI self-repair. Forestry historical holdout: eight cases with a modest improvement over naive forecasting. Agriculture: eight cases where the alternate method did worse than naive. Emergency-response: nine historic timestamp/schema cases, not operational incident truth validation.

Production status: HOLD. No completed 25-task autonomous invention qualification, no real hosted cross-model research qualification, no three independent attested physical FarmNodes, and no owner-signed ENTITY production or investment-issuance permission. Registry type designs are not deployed machines.

Council qualification report (18:26 UTC) → · PDF → · Research sector evaluation report → · Sector PDF →