Market/economic correctness
Do venues, rights instruments, balances, execution, clearing, entitlements, usage, revenue rules and settlement evidence behave deterministically and fail closed?
ENTITY research should ask falsifiable questions across the whole fabric: can unrelated implementers reproduce semantics, can rights markets preserve authority and evidence boundaries, can economic state survive provider failure, can settlement claims remain externally verifiable, and can BTDU reduce representation/infrastructure cost under independently controlled workloads?
Do venues, rights instruments, balances, execution, clearing, entitlements, usage, revenue rules and settlement evidence behave deterministically and fail closed?
Can an unrelated engineer reproduce the required ENTITY semantics from public material without BTG implementation code?
Under what workloads do reusable atoms/bonds reduce representation storage, RAM, compute or model requirements—and where do they not?
Test whether explicit rights and instrument terms can support price discovery and transfer without conflating byte possession, provenance, legal title or external payment.
Market fabric →Evaluate authorized settlement-verifier models, payment attestations, dispute/failure behavior and the boundary between execution, obligation, externally verified settlement and realized value.
Settlement docs →Test semantic snapshot/restore of venues, balances, orders, trades, entitlements, usage, revenue rules, treasuries and obligations across different storage/providers.
Recovery architecture →Study contribution instruments, accepted upstream lineage, sponsorship, service economics and downstream causal/economic relationships without retroactively changing open-source rights.
Vector & Memnox lineage →In one BTG-controlled comparison involving roughly 50 million entities and 250 million bonds, the tested BTDU representation was about 4.4 GB versus about 153.6 GB for the compared 768-dimensional float32 vector representation—about 97.14% lower representation storage in that specific setup.
This compares representation storage only; it does not prove universal gains for every database, model, RAM or compute workload.
Can the architecture sustain hundreds of millions or billions of useful relationships while materially reducing infrastructure costs across realistic workloads selected by independent researchers?
Formalize delegation, expiry, revocation, recovery and privilege non-expansion across user, organization, AI-agent and market actions.
Analyze canonical signed intents, nonces, cancellation ordering, RFQ expiry, surveillance state and duplicate/replay failure modes.
Study vector-clock/causal conflict behavior for disconnected market/economic nodes without silent last-writer-wins resolution.
Test suite-version migration, dual-sign transitions, retirement, downgrade resistance and historical-verification continuity.
Vector #26501 / PR #26504 and Memnox #46 / PR #86 are merged upstream and recorded as UPSTREAM_ACCEPTED_RECORDED. Both are pro bono with USD 0 realized cash, making them useful research cases for external contribution lineage without a confounding paid-settlement claim.