2026-07-16
Released Signal Lab method signal-lab-2026-07-16.2. Market State now uses the same two directional inputs for every asset: trend at 62.5% and funding positioning at 37.5%. Positive discovery remains visible context but no longer leaves nearly every asset unassessed merely because it is absent from a small trending list. The public cards call the ordering value a board score and retain the restored circular score dial.
Added the first owned market verification lane for fixed-block Bitcoin and ERC-20 supply receipts plus identity-matched on-chain venue reconciliation. It uses capped immutable storage, last-good preservation, two-venue agreement, freshness and dispersion gates, and explicit total-versus-circulating supply boundaries. Coinbase, OKX, and Coin Metrics Community data remain disabled where reviewed terms do not clear commercial fallback use. Owned pool volume cannot enter the live board until it is comparable with global reported volume.
Released Signal Lab method signal-lab-2026-07-16.1. Board rank now covers every asset with both required board-score inputs and remains independent from Risk Gate eligibility. An ineligible configured anchor opens another rotating slot, so a complete selected board still contains eleven assets. The public board now presents one ranked set without internal role labels or repeated placeholder values.
Added reviewed chain-slug relationships for ambiguous DeFiLlama identities, fresh dated weekly TVL comparisons, and economic materiality before a TVL change can hard-block selection. Exact protocol-family exploit matching now replaces chain-wide incident attribution, returned funds reduce unresolved loss, and malformed or short exploit ledgers fail closed. Detail hero values now carry direct source and timing receipts. Complete boards also require current full-snapshot receipts from Hyperliquid perps and Binance USD-M funding, with failed refresh status preserved on the last-complete fallback.
Tightened complete-board coverage so every monitored asset needs a live CoinGecko market row and both finite board-score inputs. DeFiLlama chains, protocols, open interest, and hacks now pass only current validated rows into Signal Lab and carry served and attempted receipts through fallback. Chain TVL history now has an explicit live, stale, or unavailable receipt, and only live history can alter the Risk Gate. Optional stablecoin, fee, revenue, and DEX fallbacks remain outside current scoring.
2026-07-15
Released Signal Lab method signal-lab-2026-07-15.2. Structural Depth, Gem Score, Market State, Risk Gate, and Evidence Grade now answer separate questions. Structural Depth and Gem Score remain withheld until their direct evidence gates exist. Missing inputs remain null and are omitted, source coverage stays outside asset arithmetic, and all-time-high distance is drawdown context rather than valuation.
Added explicit asset archetypes and separated board rank from the continuity-and-rotation selection policy. The board score is a disclosed continuity aid, not an asset-quality or hidden-gem score. Complete board snapshots now fail closed when eligibility, raw ranks, selected roles, or policy reasons do not reconcile. Daily moat vintages retain the new method, separated outputs, selection fields, and a versioned compact receipt of every normalized formula input used at scoring time.
2026-07-13
Released the first Bathymark Research Observatory as a manually reviewed, version-controlled evidence map. Thirteen source-located claims across twelve works and five research domains now keep the literature record, Bathymark reproduction state, and live-validity state on separate axes. The release reports zero Bathymark replications and zero live monitors. Source review is not presented as proof that a result survives current data, costs, venues, or market regimes.
Added a public research graph, claim sitemap, natural-language route search, and claim-level source, method, structured result, limitation, falsifier, contradiction, and status-history records. The graph exports Bathymark's curated bibliographic facts and paraphrases only. It stores no paper abstract or full text and advertises no research API, model, replication package, or agent action that does not exist.
Added a bounded owner-run Crossref metadata audit with a production source-rights gate before retrieval. Its contract permits bibliographic and registered-update metadata only. Abstracts, publisher text, excerpts, automated claim extraction, and model processing remain outside the approved source asset. OpenAlex remains an unconfigured expansion source pending a separate rights and cost decision.
2026-07-12
Added the versioned SOUNDING evidence workflow, 18 registry-driven Claim Checks, and the browser-local Decision Ledger. Claim Checks separate the recorded statement source, dated review scope, document-linked notes, dependency provenance, missing proof, hard limits, and review triggers. A missing statement remains a coverage gap instead of becoming an issuer claim. Decision records stay on the user's device, make no model or server request, synchronize same-origin tabs, preserve the original precommitment, and accept one immutable outcome.
Expanded Bathymark Academy from eight pathways and 32 lessons to twelve pathways and 44 lessons. The original curriculum remains intact, with new Bitcoin evidence, custody, exposure and exit, and decision-operations paths. Every pathway and lesson now maps to the same SOUNDING steps and retains its source pack, limitation, project, and deterministic self-check.
Published the human Trust Center and Trust Manifest version 1.1, plus a dedicated Claim Check sitemap and bounded public route search. Retrieval agents can query public route metadata while private API surfaces remain blocked. The source registry now separates public access from commercial, caching, redistribution, adaptation, export, and model-training rights. Unknown, pending, expired, and runtime-invalid contracts approve no new use. New adapters must invoke the gate, and the public registry updates its effective review state daily.
Added a source-backed UTC month chart to Tide Map and a linked evidence-path visual to Claim Check. Calendar cell glow shows event count only, pressure colour repeats the published categorical state, and neither visual claims direction or probability. Full evidence cards, tables, source receipts, and mobile agendas remain available in visible HTML.
Added a versioned product evidence coverage engine and terminal audit. The RWA hub now consolidates unresolved product-level exit depth into one honest coverage state, reuses current fixed-block Exit Sounding receipts when they exist, omits empty comparison columns, and exposes the source, control, redemption, custody, NAV, and exit fields Bathymark should acquire next. The queue is deterministic and needs no model or paid API key.
Released Signal Lab method version signal-lab-2026-07-12.1. The 200-day moving average now requires 200 consecutive UTC-day observations, rejects future and stale endpoints, and keeps its new vintages separate from the previous method. The detail view requests a longer bounded primary-asset history so the rolling line retains about 200 plotted values without enlarging the full 88-asset refresh.
Corrected Overlay Lab comparison math. Moving series now share one real date window and each begins at index 100, while fixed price references retain the asset-price denominator needed to show their actual level. Missing dates are not joined, disconnected ranges are withheld, current heuristic references are labelled as unrated, and quote time remains separate from the shared-window daily close.
The bounded history loader now attempts every configured Signal Lab asset after an empty intermediate batch. A partial or failed attempt still cannot replace the last complete 88-asset board, and stale or partial screens no longer imply a current scenario band or ranking. Price-history fallback status, capture time, check time, and limitation now travel into the selected asset and benchmark source receipts.
2026-07-11
Released Waterline 2.0 with confidence kept separate from market state, versioned metric definitions, separate source-observation and snapshot-capture receipts, fail-closed coverage, and a stored-history gate for persistence and percentile claims. Nominal USD TVL changes are now labelled as nominal depth rather than deposits, withdrawals, or fresh capital. The value and availability gate use five weighted source lanes: protocol and chain TVL, stablecoin references, paid-fee activity, and open interest. DEX turnover and yield screens are preserved as separate context in The Current and do not enter the index.
Added the stablecoin instrument taxonomy and corrected the current list metric to source-computed circulating USD market value, while keeping per-token history in token units. Variable-price accumulating-NAV units no longer receive a mechanical one-dollar verdict. Fixed-price rebasing units remain eligible, non-USD price references require FX, Brazilian real peg types map to BRL, and variable targets never enter the FX path. The affected interpretation is recorded on the public corrections page.
Added the shared seven-block analysis format, Macro Liquidity Overlay, material-change watchlist, Depth Check distribution packet, resolution receipts in The Log, question-shaped research hubs, and an explicitly separate noindex After Dark culture shelf. New numerical indices remain frozen while longitudinal evidence matures.
Renamed the RWA Source Freshness Index display to RWA Source Trail Coverage. The existing route remains available, while the reading now states its real scope: named source and URL coverage, not reachability or publication-age freshness.
Added the Intelligence Graph, The Helm, Notice to Mariners, Oracle Watch, Exit Sounding, Depeg Memory, and Yield Provenance as additive product-trust layers. Public product passports now keep control, dependency, change, valuation input, redemption, and observable exit evidence beside the existing backing and source trail. Missing control or exit evidence remains unresolved and never becomes an immutability or safety claim.
Hardened Signal Lab refresh behavior for its 88-asset universe. Only a structurally complete board may replace the last complete ranking, while a partial attempt keeps all asset evidence routes available and publishes its own coverage receipt. A stale complete board now identifies both the served snapshot and any newer failed attempt instead of presenting an incomplete refresh as a healthy ranking.
Added the categorical Tide Map over configured token unlock, official BLS, reviewed FOMC, and BTC and ETH dated-futures lanes. Governance and protocol upgrades remain visibly unconfigured. Tide Pressure stays separate from source Visibility, and the immutable Evidence Ledger preserves original Pressure Window publications, later outcomes, and additive corrections. Calibration remains withheld without comparable precommitted binary results.
Opened the Bathymark Academy self-guided beta with eight project pathways, fixed self-checks, browser-local progress, and a deterministic citation-first study desk. Live cohorts, grading, certification, and generative tutoring remain planned features rather than current product claims.
2026-06-27
Added query-shaped SEO and GEO surfaces: grouped sitemaps, answer blocks, search hub pages, expanded entity templates, richer glossary pages, trust pages, and clearer metadata for core liquidity routes.
How updates are recorded
Material method and source changes should be recorded here or in Methodology. Routine copy edits and visual refinements do not need a log entry unless they change the meaning of a reading.