Yield Intelligence Changelog

Impact Notes

• `usdai-usd-ai` is no longer treated as a yield-bearing wrapper host, so base USDai stops inheriting sUSDai's savings pool as its native yield source
• New tracked asset `susdai-usd-ai` now owns the existing USD.AI savings pool mapping directly, aligning yield rankings with the dedicated sUSDai detail page and nav-token semantics
• The stale USD.AI wrapper config that pointed at Arbitrum PYUSD as the `sUSDai` variant address is removed, eliminating a concrete address-level misbinding

Impact Notes

• `bd-basedollar` is now classified the same way as other pre-launch yield-bearing assets with no live runtime source, so the manifest no longer reports it as a covered entry with zero strategies
• Coverage audits and operator tooling continue to inventory every yield-bearing asset, but pre-launch gaps are now fail-closed and explicit rather than implicit
• The public yield methodology docs and timeline now call out the intentional-gap treatment for these pre-launch assets

Impact Notes

• For tracked stablecoins, lending-opportunity rows now require `sourceTvlUsd` and must clear `max(existing absolute floor, 0.1% of current supply)` before publication
• This applies across auto-discovered DeFiLlama lending markets, deterministic exact-pool overrides, and supplemental protocol-native lending venues
• TVL-less protocol suggestions no longer fail open into live recommendations; until venue size is observable they are omitted from published lending-opportunity coverage
• Yield methodology docs and changelog entries now document the new supply-relative recommendation gate explicitly

Impact Notes

• The supplemental Yearn/Kong reader now recognizes Ethereum `Staked yBOLD` and pins it to `bold-liquity` as `K3: sBOLD`
• This source is classified as `lending-vault`, keeping BOLD's wrapper-over-wrapper Stability Pool path in the native-yield bucket rather than `lending-opportunity` or `governance-set`
• Source-link resolution now deep-links `K3: sBOLD` to Liquity's dedicated earn route
• Yield methodology docs and the public changelog now document the additional native BOLD source coverage

Impact Notes

• Supplemental protocol-API lending candidates such as `Morpho: Resolv USDC` are dropped before ranking publication when the venue label resolves to Resolv / `USR`, `stUSR`, or `wstUSR` exposures
• Auto-discovered DeFiLlama lending pools now preserve `poolMeta` in the shared cache and apply the same Resolv / USR exclusion rule, keeping the hourly publisher and the slower supplemental lane aligned
• The exclusion is scoped to `lending-opportunity` venues, so native tracked yield assets and their own methodology coverage remain unchanged
• Wrapper-over-native venues such as BOLD / `yBOLD` are documented and classified as native yield rather than `governance-set` when the wrapper only packages the protocol's own Stability Pool return
• Yield methodology docs and the public changelog now document the explicit USR-linked venue exclusion rule

Impact Notes

• PYS now computes `effectiveYield = max(0, apy30d + 0.25 * (apy30d - benchmarkRate))` before dividing by the adjusted risk penalty, so local-currency benchmark outperformance affects the score directly
• The change rewards rows that clear tighter EUR, CHF, or other native hurdles without turning PYS into a pure excess-yield ranker
• Worker-time scoring and live `/api/yield-rankings` safety hydration now pass row benchmark context into the shared scorer, removing another source of score drift risk
• Leaderboard/detail breakdowns, methodology docs, and yield-changelog entries now expose the benchmark adjustment and effective-yield terms explicitly

Impact Notes

• Supplemental-backed protocol-API rows now wait 6 hours before surfacing `data-stale`, so normal end-of-cycle hourly publishes no longer show false stale warnings
• Optional Aave V3 and Compound V3 rows now use the same 6 hour freshness window because they are refreshed by `sync-yield-supplemental`, not the hourly publisher
• Deterministic hourly on-chain rows keep the existing three-hour stale threshold, so only the slower supplemental families move
• Yield methodology and operations docs now distinguish hourly, supplemental, and daily freshness windows explicitly

Impact Notes

• Compound V3 now probes both configured RPC endpoints instead of only the fallback URL, and Aave V3 plus Compound V3 rotate endpoint order across targets with a slightly deeper retry budget on the best-effort supplemental lane
• `sync-yield-supplemental` metadata now records optional RPC family target counts, attempted counts, resolved target counts, emitted row counts, missing target counts, per-chain miss breakdowns, and miss reasons
• Layer 2 wrapper matching can now pin a preferred DeFiLlama project in addition to chain and address, so shared wrapper tokens like `sUSDe`, `sUSDS`, and similar cases stay fail-closed without attaching to the wrong venue
• Under-specified wrapper configs now carry explicit live chain/address/project pins for native venues such as `sUSDai`, `sNUSD`, `savUSD`, `sUSDu`, `syzUSD`, `sAID`, `stCUSD`, and `sGHO`

Impact Notes

• Aave V3 supplemental supply-rate rows now participate in arbitration as curated protocol-native venues rather than deterministic wrapper rows
• Native wrapper yields such as sDAI no longer lose the primary row to a lower-yield supplemental lending market purely because the lending reader used an on-chain transport
• Source keys and alternative-source history remain unchanged, so only arbitration precedence moves
• Yield methodology docs and the timeline now document that Tier 2.5 lending readers do not outrank stronger native wrapper yields by source family alone

Impact Notes

• Mixed-benchmark scopes such as the default `All` view now render the horizontal benchmark line and four shaded quadrants again
• When the visible set mixes USD, EUR, and CHF hurdles, the scatter frame explicitly uses the USD benchmark as a shared visual reference
• Mixed-view copy now tells users that the background zones are an orientation frame and that each row's benchmark tag still governs excess-yield interpretation
• Yield methodology docs and the changelog now describe the restored mixed-view scatter behavior explicitly

Impact Notes

• `price-derived` rankings now use a 36 hour stale threshold because they are backed by daily `supply_history` snapshots rather than hourly source observations
• Hourly publication families such as on-chain, DeFiLlama, protocol-native, and auto-discovered rows still mark stale after three missed `sync-yield-data` intervals
• Healthy daily snapshot rows such as USTB, USDA, and CETES no longer show false `data-stale` warnings after roughly one day of normal operation
• The methodology docs, changelog, and operations note now document the source-cadence freshness windows explicitly

Impact Notes

• EUR benchmark refreshes now use the ECB's official 3-month compounded €STR series (`EST/B.EU000A2QQF32.CR`) instead of the overnight €STR series
• CHF benchmark refreshes now fetch delayed public `SAR3MC` from SIX via the guest OAuth plus report-download flow, replacing the old SNB policy-rate proxy
• CHF benchmark entries are no longer labeled as proxies, and mixed-benchmark UI copy now names the 3-month compounded EUR and CHF cash hurdles explicitly
• Methodology docs, API examples, and the about-page source inventory now reflect the new EUR/CHF risk-free benchmark stack

Impact Notes

• Aave V3 supplemental on-chain rows now use asset-scoped source keys, so same-chain markets no longer overwrite each other in the supplemental cache
• `sync-yield-supplemental` now reports raw candidate count, deduped candidate count, and dropped-row count in cron metadata so silent row loss becomes visible to operators
• The monthly coverage audit now counts explicit auto-discovery overrides and curated exact-pool overrides as covered DL surfaces instead of treating only `YIELD_POOL_MAP` UUIDs as covered
• High-TVL coverage-gap reporting now focuses on unsupported protocol surfaces rather than flooding the audit with already-supported allowlisted markets

Impact Notes

• `data-stale` now triggers after three `sync-yield-data` intervals (currently about 3 hours) instead of after a hard-coded 90 minutes
• Stablecoin detail Yield Intelligence cards no longer flag healthy hourly snapshots as stale after only one delayed publish window
• The stale-warning threshold is now derived from shared cron metadata so future cadence moves stay aligned automatically
• Methodology docs and the internal timeline now describe the cadence-aware stale-warning rule explicitly

Impact Notes

• EUR benchmark refreshes now try the ECB Data API first and only fall back to the FRED €STR mirror when the official feed is unavailable
• CHF benchmark parsing now normalizes the SNB page to text before extracting the policy-rate sentence, avoiding breakage from harmless markup changes
• Degraded benchmark metadata now reports explicit EUR and CHF failure modes instead of collapsing first-run failures into a generic `unavailable` bucket
• Yield methodology, API examples, and source inventory now reflect the official ECB feed plus the hardened SNB proxy parser

Impact Notes

• PYS now computes yield efficiency as `apy30d / (riskPenalty ^ 1.75)` instead of dividing by a linear safety penalty
• The global `PYS_SCALING_FACTOR` increased from `5` to `8` so score distribution remains readable after the steeper safety curve
• Live `/api/yield-rankings` hydration now reuses the shared PYS scorer, removing formula drift risk between cron-time scoring and read-time safety hydration
• Leaderboard, detail-surface breakdowns, docs, and the methodology page now reference the adjusted risk penalty explicitly

Impact Notes

• The benchmark cache now publishes a small benchmark registry instead of only a single global USD risk-free rate
• Each ranking row now exposes its selected benchmark label, rate, fallback state, and selection mode so excess-yield semantics remain explicit downstream
• Detail pages, hero chips, and yield-history charts now label excess yield against the row's actual benchmark instead of hard-coding `vs T-Bill`
• The `/yield` page now hides the single benchmark line on mixed-benchmark views and restores it only when the visible scope shares one benchmark
• CHF support uses the public SNB policy rate as a proxy rather than the SNB-published SARON display, whose usage is restricted

Impact Notes

• The `/yield` page now supports peg-scoped ranking views, including a `non-usd` preset that groups the live EUR, CHF, SGD, MXN, and commodity rows into one visible universe
• Tier-2 DeFiLlama ingestion now preserves exact curated non-stablecoin pool UUIDs in addition to native pool IDs and wrapper symbols
• A new exact-pool override lane can publish assets like `xaut-tether` from a named venue when the UUID, project, chain, and symbol all match and the APY/TVL quality gates pass
• Generic gold/silver auto-discovery remains disabled, preventing mixed baskets such as Multipli's RWAUSDI pool from being misclassified as single-asset commodity yield sources

Impact Notes

• DeFiLlama discovery, variant matching, and protocol-native adapters now prefer chain+address identity and drop ambiguous symbol-only candidates instead of attaching them to the first matching coin
• Protocol-native source keys now use full chain-aware identifiers (Morpho, Pendle, Yearn, Kong, Beefy, Compound, Aave) and source-link matching understands prefixed and chain-qualified labels
• Yield manifest coverage now includes explicit price-derived fallbacks and intentional gaps, so assets like cetes-etherfuse and usg-tangent are no longer invisible to coverage reporting
• Warning heuristics and published `medianApy` now share the same TVL-weighted 30d benchmark
• yield-history no longer advances past the latest published yield-rankings snapshot when DB writes and cache publication diverge

Impact Notes

• Chain-scoped matching adds optional chainFilter to findBestLendingPool, derived from coin contract deployments
• Variant scanner detects sXXX/stXXX/wXXX prefix and SAVE/VAULT/EARN/STAKE suffix patterns in DL pools
• Monthly coverage audit cron (1st of month, 06:00 UTC) flags unmatched high-TVL pools and missing protocols
• Protocol recommendations classify missing protocols as high-confidence (>$10M, 3+ pools) or review-needed

Impact Notes

• 10 protocol-native adapters provide direct yield data, reducing DeFiLlama intermediation
• Aave V3 + Compound V3 direct on-chain supply rates across Ethereum, Arbitrum, and Base
• Kong adapter covers 2,083 ERC-4626 vaults (Yearn + Morpho + Spark + Fluid + others)
• USTB + thBILL upgraded from T-bill proxy to actual on-chain ERC-4626 exchange rate reads
• DeFiLlama yield history backfill gives instant 365-day charts for newly tracked coins
• Expanded lending protocol allowlist with 19 new protocols (Wildcat $235M, Tectonic $100M, etc.)
• cusd-cap flagged as yield-bearing with stCUSD savings wrapper ($68M TVL)
• Lower TVL floor ($25K) captures Solana/Sui/Aptos/Cardano/Stacks lending markets

Impact Notes

• usbd-bima now emits a protocol-native `BIMA savings (sUSBD)` source row sourced from BIMA's published `/api/earn/pools` feed
• Yield arbitration treats protocol-owned earn APIs as curated sources rather than misclassifying them as on-chain or DeFiLlama data
• The about page and source-link registry now expose BIMA's earn surface as an official yield-source reference

Impact Notes

• Price-derived and on-chain rankings now expose the age of their actual observation inputs instead of always presenting as fresh at sync time
• Derived-source provenance now includes comparison-anchor timing so older anchors are visible to downstream consumers
• All allowlisted lending protocols now resolve to curated source links instead of falling back to coin-level websites or nulls

Impact Notes

• Rankings publication is now validated before live row mutation, reducing DB/cache divergence risk on schema or publish-guard failures
• Degraded runs retain prior current rows by skipping destructive cleanup instead of deleting rows while upstream inputs are impaired
• dUSD and reUSD were removed from the generic ERC-4626 deterministic adapter set until protocol-specific readers exist, reducing false confidence in Tier 1 coverage

Impact Notes

• Polaris pUSD now resolves through a deterministic Silo v2 lending market override, fixing the prior bypass-only configuration gap
• USDX, USDO, and USDM now use deterministic exact-symbol lending overrides rather than waiting on the generic dynamic discovery path
• These explicit overrides bypass the normal C- safety gate only for a short named list of high-TVL or protocol-native edge cases, preserving the broader global discovery standard

Impact Notes

• Auto-discovered lending opportunities now accept single-asset pools down to $100K TVL and 0.10% APY, capturing still-meaningful long-tail markets without opening full dust-pool coverage
• The curated lending allowlist now includes More Markets and SmarDex USDN, while Polaris pUSD can bypass the report-card gate through an explicit vetted edge-case override
• Price-derived APY now annualizes from the oldest available price anchor between 7 and 45 days instead of requiring a strict 30-day sample, improving early coverage for newer NAV tokens

Impact Notes

• USYC and thBILL now participate in rate-derived Treasury fallback coverage alongside the existing BUIDL/USTB/YLDS/mTBILL/OUSG set
• Auto-discovered lending coverage now recognizes additional curated protocol slugs already present in live DeFiLlama data, including Loopscale, Vesper, Lista Lending, Liqwid, Overnight, Lagoon, and NAVI Lending
• The lending auto-discovery TVL floor was reduced from $1.0M to $0.5M to capture still-meaningful long-tail lending markets without opening the door to dust pools

Impact Notes

• Direct DeFiLlama yield fetches now degrade on invalid payloads or zero relevant stablecoin pools instead of being treated as a benign empty set
• Runs now degrade when all configured deterministic on-chain sources fail in the same cycle, exposing that outage in rankings provenance
• Retained `risk_free_rate` fallbacks preserve the last market-derived benchmark fields across degraded streaks, and rankings-cache publication now blocks on severe shrink versus the prior cache

Impact Notes

• On-chain rate configs now emit a seed row when no previous exchange rate exists, breaking the bootstrapping deadlock that prevented Tier 1 APY computation since launch
• buildOnChainSourceKey consolidated from 3 duplicate definitions into a single shared export
• Pool pre-filter set allocations promoted from per-call to module-level for improved DL pool ingestion performance
• Live safety hydration coverage ratio now uses active card count instead of total card count for accurate degradation detection

Impact Notes

• DeFiLlama pool ingestion now retains single-exposure wrapper pools that are explicitly relevant via native or variant config even when upstream `stablecoin` flags are false
• Deterministic on-chain rows now use `onchain:<stablecoinId>` source keys so previous-rate lookups and source-aware history cannot collide with curated pool UUIDs
• Live `/api/yield-rankings` safety hydration keeps rows with `DEFAULT_SAFETY_SCORE` / `NR` when report-card coverage is incomplete instead of dropping yield coverage
• Retained benchmark fallbacks stay marked as degraded, and malformed stored `warning_signals` payloads no longer fail `yield-history` requests

Impact Notes

• yield_history now persists per-source rows with best-source markers instead of a single mixed best series
• 7d and 30d APY metrics are computed from source-specific history, preventing source-switch contamination
• Rankings now include provenance for benchmark freshness, safety coverage, source-switch state, and selection reasoning
• Cross-source arbitration can reject divergent discovered or fallback sources when canonical sources disagree materially

Impact Notes

• Added direct on-chain LUSD source using Liquity Stability Pool deposits and CommunityIssuance totals
• APR converts projected LQTY emissions to USD using CoinGecko spot price and excludes ETH liquidation gains by design
• LUSD can now surface both B.Protocol Stability Pool and auto-discovered lending alternatives in the same ranking payload

Impact Notes

• yield_data primary key changed to (stablecoin_id, source_key) with per-source rows
• is_best now marks the highest-APY source per coin; non-best alternatives are retained
• Tier 2 matching aggregates all valid sources (native map, wrapper map, symbol fallback)
• /api/yield-rankings now includes altSources[] and UI exposes +N alternative source details

Impact Notes

• Auto-discovery added minimum APY/TVL filters and expanded protocol allowlist coverage
• Deterministic pool overrides introduced for hard-to-match symbols (including explicit safety bypass handling)
• findBestLendingPool now falls back to underlying token address matches when symbol matching fails
• Price-derived fallback explicitly extended to BUIDL when no usable on-chain or DL source exists

Impact Notes

• Resilience inputs in inline safety computation now use shared isBlacklistable() logic
• Risk penalties in PYS better reflect inherited blacklist exposure
• Reduced divergence between yield-page safety grades and safety-scores page outputs

Impact Notes

• NAV tokens were included in inline safety scoring instead of defaulting to implicit NR behavior
• Yield sync now reuses cached DeFiLlama pools from DEX sync to reduce upstream fetch failures
• Non-finite 30-day APY volatility values are sanitized before D1 writes

Impact Notes

• Added allowlist-based auto-discovery pass over DeFiLlama lending pools
• Eligibility gated by safety score threshold before pool selection
• Introduced defillama-auto source type and lending-opportunity yield classification

Impact Notes

• warning_signals persistence added with spike/divergence/trend/reward/TVL-outflow checks
• Signal detection now uses market-median APY and prior TVL context per coin
• Tier-2 deterministic source map added explicit fxUSD Stability Pool coverage

Impact Notes

• Added wave-1 variant/pool mappings for additional native-yield stablecoins
• Near-zero mean handling in stability/variance math prevents coefficient-of-variation blowups
• Safety fallback and finite-value guards were formalized for ranking writes

Impact Notes

• 7-day APY switched to timestamp-window filtering instead of proportional sample slicing
• Tier-1 previous exchange-rate reads were reused from cached lookup state
• Yield stability display normalized as a true 0-100 percentage in UI components

Impact Notes

• Introduced three-tier APY resolution (on-chain rate, DeFiLlama pool, NAV price-derived fallback)
• Launched PYS model (risk penalty + variance sustainability multiplier + scaling factor)
• Added yield_data/yield_history tables and public yield-rankings/yield-history API handlers