Every research update we ship, logged and classified. 73 updates across 14 projects over 8 days.
Previously the worker fetched the previous snapshot from the static /data/leverage-audit-latest.json, which only updated when a human ran the CLI locally and committed. This commit adds Supabase persistence so the diff auto-updates each run.
One-shot fix sweep driven by the content-leverage-audit output. Takes the audit from 81 total violations to 0 and from 5 cross-protocol awareness gaps to 0 in a single commit.
Two bugs in renderScatter() the expert review caught:
New concept page segmenting onchain lending into three architectures and matching each to the risk profile it's best suited for. Monolithic pools (Aave, Spark, Compound) for continuous price risk. Modular (Morpho, Euler, Kamino) for hybrid/jump risk. CeDeFi (Maple, Pareto, Wildcat) for credit risk that can't live in permissionless systems. April 2026 Kelp/Aave incident treated as the cleanest worked example of architectural trade-offs.
The methodology page now uses Anastasiia's Layer 1 (mechanical) / Layer 2 (governance) / Layer 3 (code integrity) taxonomy, with the dominance condition (Layer 3 must be cleared before Layer 1 optimization is meaningful). Six failure modes documented where TradFi credit analogies break in DeFi. Resolv reference case added as a worked example of oracle latency.
New curator-layer section on the Morpho research page quantifying where the protocol's risk actually sits. ~$7.27B across 8 major curators (Gauntlet 27.6%, Steakhouse 17.8%, MEV Capital 12.6%, K3 Capital 6.6%, mid-tier ~35.4%). Network centrality finding: B Protocol and Block Analitica are the ecosystem hubs (eigenvector centrality ~0.46 and ~0.43) despite being mid-sized. Lower-tail drawdown correlations stay elevated where collateral overlaps.
The solver-landscape concept page now covers three solver categories: market-maker solvers, algorithmic solvers, and the new Solana private-DEX tier. The taxonomy is useful for understanding where DEX fee capture actually accrues.
Jump Crypto data documented: BisonFi, HumidiFi, SolFi, Obric v2, and ZeroFi together processed ~$19.87B PropAMM volume in March 2026, matching the top-4-CEX aggregate on SOL dollar pairs. Toxic-flow dynamics mean the public AMM layer (Raydium, Orca) increasingly receives residual flow. This is the Solana variant of the Uniswap fee-switch problem and worth tracking for RAY / JUP revenue durability.
Explicit documentation of the April 2026 Kelp DAO LayerZero bridge exploit ($292M drained) and its second-order impact on Aave: $6.6B Aave TVL exited in 24 hours, $196M in bad debt accrued against the Umbrella reserve module. The risk-framework accountability angle: 98.5% of WETH collateral was LST-concentrated and the pool had priced continuous price risk correctly but not LST-correlated jump risk.
The cross-chain bridge configuration (DVN count, messaging-layer guarantees) becomes an explicit additional check in the TokenIntel DeFi Risk framework. Kelp DAO's 1-of-1 DVN setup becomes the cleanest reference case for why single-verifier bridge configurations are a structural vulnerability.
Dedicated 'Toxic Flow Risk from Intent-Based Routing' subsection added to the Uniswap research page. Scored medium / structural / uncertain magnitude. UNI's thesis depends on fee-switch revenue quality, which depends on what intent-based solvers are pulling out of the AMM order flow. Multi-primitive risks noted alongside.
New concept page documenting how market-maker solvers, algorithmic solvers, and the new Solana private-DEX tier determine toxic flow, AMM LP economics, and where fee switches actually capture value. Framework for understanding UNI fee-switch quality and PropAMM dynamics.
All 6 AI content generators (weekly newsletter, social growth, social content, monthly recap, crypto digest, copilot daily brief) now flag violations instead of killing outputs. Review banners + admin alerts + option-2 fail-safe dashboard fallback. Reduces false-negative content loss while keeping accuracy guardrails active.
Frontend now uses CMC's Fear & Greed index proxy instead of alternative.me. Fixes the root cause of a March 2026 incident where stale content generators cited alternative.me's Extreme Fear reading (16) while CMC showed Neutral (49), producing conflicting narratives in subscriber emails.
Multiple stale claims on BTC, ETH, and XRP research pages corrected against current audit data.
Uniswap research page now distinguishes between eligible market cap (the ERM denominator) and protocol TVL (a separate operational figure) to prevent false staleness-audit flags and accidental comparison mistakes.
Chainlink research page clarified that the $95B figure cited is cumulative (lifetime) Total Value Secured, not a current TVS snapshot. Sourcing added to prevent future misreads.
The admin-override path for TokenIntel signals has been removed. target_signal logic retained for alerting, but the algorithm is always the source of truth on the dashboard. Reduces editorial-vs-algorithm drift.
Raydium research page refreshed with current Q1 2026 fundamentals from Token Terminal.
The April 2026 SEC staff guidance that non-custodial crypto interfaces can operate without broker-dealer registration was integrated into the Regulatory Risk sections of 5 DeFi research pages (Aave, Morpho, Maker, Maple, Ethena). Caveats on permissionless-market and integration-with-custodial-venues edge cases noted.