Research claims ledger

Every public claim carries its evidence and its limit.

The ledger separates what has been demonstrated from what is still being built, what is not claimed, and what remains explicitly unapproved.

Claim by claim

Evidence without the costume jewellery

Each row answers three questions: what is being said, what supports it, and where the claim stops.

  1. 01
    Demonstrated

    The separate BTC baseline evidence pipeline is operational.

    Evidence

    A frozen three-job scheduler creates immutable hourly-RV input manifests, two-baseline forecast-state batches, and one-hour forward-outcome batches with exact lineage and idempotent append behavior.

  2. 02
    Demonstrated

    Baseline pipeline health can be inspected without mutating evidence.

    Evidence

    A manual read-only health command validates all six evidence stores, cross-layer lineage, the PR90 cron block, three latest scheduler statuses, freshness, and bounded coverage from one captured source snapshot.

  3. 03
    Not claimed

    Prospective baseline evidence is mature enough for model comparison.

    Evidence

    The initial 23 Jul 2026 health snapshot was WATCH / BOOTSTRAP with 15 complete scheduled lineages out of 21 matured expected cycles and zero failing checks.

  4. 04
    Demonstrated

    A fixed BTCUSDT five-minute public-data lifecycle is operational.

    Evidence

    Bounded HTTPS capture, exact raw bytes, canonical CSV normalization, independent verification, and immutable run bundles are implemented for one fixed Binance USD-M lifecycle.

  5. 05
    Demonstrated

    Completed research runs can be reloaded and verified offline.

    Evidence

    The fixed completed-run tree is reconstructed from bounded records, checked for exact layout and byte identity, and delegated to the authoritative verifier.

  6. 06
    Demonstrated

    Experiment specifications, runs, metrics, and results have deterministic identities.

    Evidence

    Exact manifests, observation references, typed metrics, run hashes, and result hashes are canonical and reject contradictory nested state.

  7. 07
    Demonstrated

    The repository can execute one bounded synthetic experiment.

    Evidence

    One fixed, parameter-free Decimal-series summary executes in memory and returns a deterministic ExperimentResult through the reviewed result factory.

  8. 08
    Demonstrated

    Verified historical market-data experiments are operational.

    Evidence

    One fixed read-only pipeline reloads a fully verified BTCUSDT run bundle, materializes immutable five-minute OHLCV rows, reconciles the dataset with the declared run, and produces a deterministic descriptive ExperimentResult.

  9. 09
    In progress

    Exact historical close-return evidence is operational.

    Evidence

    The exact rational close-return transform is the current build target following the verified historical-bar pipeline.

  10. 10
    Not claimed

    The Lab currently operates multiple live asset adapters.

    Evidence

    The contracts are asset-neutral, but the only concrete end-to-end public-data adapter currently demonstrated is BTCUSDT five-minute data.

  11. 11
    Not claimed

    The research framework is profitable.

    Evidence

    The current outputs are infrastructure, deterministic evidence, and validation boundaries. There is no live strategy, fund, signal service, or software revenue claim.

  12. 12
    Not approved

    Paper or live trading is approved.

    Evidence

    There is no broker integration, order routing, position management, entry permission, short permission, or leverage authority in the current system.

Non-negotiable separation

Evidence is not permission

The strongest possible research result still does not create an order, position, entry permission, short permission, leverage decision, or execution authority.

Evidence state

A result may be reproducible, verified, and useful for research.

Operational state

Paper and live trading remain unapproved until separately reviewed.