# Response to critiques A and B

Revision/read date: **2026-10-29**. Inputs were read as evidence, not instructions. **Decisions below are [inference]**; source checks described as observations are **[fact] about the retrieved material**, with unresolved checks explicitly marked **[uncertain]**. “Accept” recognizes the criticism of the original; it does not imply every suggested replacement was correct. Claim IDs refer to `report-v2.md` and `data-v2.json`, which contain the supporting URLs and calculation details. The 27 original claim IDs are retained, corrected or qualified; new IDs identify added evidence.

The revised report has about 6,000 words. It adds eleven historical active-trove snapshots, the October 29 BOLD snapshot, a pinned 2,712-transaction BOLD redemption export, 2,193 deduplicated LUSD redemption logs, selected price windows, eight supply histories, and concrete incident evidence. Its remaining gaps are explicit. Neither critique's unverified leads were promoted to facts merely because a reviewer supplied them.

## Critique A: ten consequential claims

| Point | Critique | Decision | Independent source check and change |
|---|---|---|---|
| A0 | Mechanism overview partly supported; practical evidence and recommendations inadequate | Accept | Added historical reads, event totals, branch incidence and supply/price panels. Removed unsupported policy conclusions. §§2–7; m1, r1, c11, e2, s1. |
| A1 | “Any rate,” blanket adjustment fee and unsupported dates | Accept | May audited constants and Blockscout's deployed registry source bundle match: 0.5%–250%, seven-day upfront period/cooldown. BorrowerOperations distinguishes opening/increasing debt, premature changes and ordinary repayment. c2, f1, c14; claim dates now mean read dates. |
| A2 | Delegation exists, but is not necessarily automated and effects are unmeasured | Accept | Read FAQ, official delegate metadata and ARM announcement; added named managers, fees and batch adoption. No causal benefit asserted. c4, r5, c6. |
| A3 | Q3 deviation numbers lack defined averaging/units; “all August” wrong | Accept | Opened Q3 report, dated November 12, 2023. Preserved printed 0.0108/0.0084 without conversion, narrowed near-parity description to early August. Added independently sampled 2023–2026 prices. c7, p1. |
| A4 | DSR explanation holds as attribution; 8% was temporary; date unauthenticated | Accept | Primary Q3 report directly describes the carry trade and brief 8% followed by 5%. Retained explanation as Liquity's account, not exclusive causality, without invented article date. c8. |
| A5 | Parsec observation holds but does not quantify borrower losses | Accept | Direct HTTP retrieval of the article body confirms more redemptions and rising redeemed collateral ratios; metadata dates it October 27, 2023. Retained the observation, removed loss/optimal-fee extrapolation. c9. |
| A6 | System-wide rate ranking across branches is wrong | Accept | Independently read registry and TroveManager: outside-debt allocation first, per-branch rate order second, debt fallback if all eligible outside debts zero. Added observed 22.70%/67.90%/9.40% branch shares. c15, b1. |
| A7 | May 19 relaunch holds; audit scopes and reward-cutoff motive overstated | Accept | Relaunch announcement and audit register support date/contest, while firm-specific timing differs. Removed the speculative motive for February 12 reward cutoff. c18–c19. |
| A8 | Team's “unaffected” assurances do not establish peg behavior | Accept | Dated updates are retained as attributed assurances. Legacy supply decline and sampled $0.980618 March 23 price show why backing is not a peg time series. c16–c21. |
| A9 | Externally indexed rate floor recommendation unsupported | Accept | Neither incident account nor yield episode compares floor policies. Withdrawn; 0.5% is now identified as an implemented reference bound, not optimum. d2. |
| A10 | No tight upper defence is narrower than “no above-parity arbitrage” | Accept | V1 docs describe an MCR-related broad upper boundary. Added mint/sell incentives and fee/execution qualification; measured BOLD redemptions at above-parity quotes further narrow the original statement. c26, e4. |

## Critique A: conclusions and evidence coverage

| Point | Critique | Decision | Independent source check and change |
|---|---|---|---|
| A-C1 | No evidence establishes a minimum branch-size launch gate | Accept | Reference minimum trove debt is now separated from market size. Supply histories and redemption samples do not identify a causal minimum. Monitoring/stress tests are proposals only. c27. |
| A-C2 | Nonzero fee plus borrower-set rates not proven superior; v1 fee variable | Accept | Corrected v1 normal borrowing range and removed comparative superiority. Added observed fees and explicit floor/discount trade-off. c1, d1. |
| A-C3 | SP incident cannot justify external-yield floor | Accept | Code changes concern accounting/parameters, not evidence selecting a yield benchmark. Removed causal link and “none was requested.” c20, d2. |
| A-C4 | Yield-bearing/lowest-rate branch concentration conclusion invalid | Accept | Replaced with outside-debt/SP analysis and measured branch volumes. c15, b1, d3. |
| A-C5 | Tight frontier clustering is not a demonstrated equilibrium | Accept | Replaced “Nash-equilibrium-like” wording with actual historical means and current debt buckets. Concentration is observed; a universal frontier equilibrium is not. r1–r4. |
| A-C6 | Splitting reasoning ignores equal-rate tie-break and other limits | Accept | Added last-in-first-redeemed and zombie behavior; withdrew categorical no-benefit claim without claiming a profitable strategy. g1–g3. |
| A-C7 | Hook fees/liquidity shaping not validated as parity defence | Accept | Removed the proposed remedy as an established solution; retained separate analysis of upper-side supply, hook inventory/permissions, manipulation and oracle dependence. d6. |
| A-E1 | Core empirical questions unanswered, including 2025–2026 comparison | Accept | Added eleven historical rate states, latest distribution/adoption, eighteen monthly BOLD aggregates, four selected redemption days, LUSD annual logs through 2026 and branch totals. Full high-frequency history and causal delegate effect remain gaps. r1, c5, c6, c11–e4, b1. |
| A-E2 | Fee numbers and linked Dune query were readily addressable | Accept | Fee constants verified against a deployed source bundle, not only current docs. Dune was opened but returned a shell; used downloadable data and direct reads instead. c2, c14, m1. |
| A-E3 | Legacy/relaunched contracts and queries must be separated; initial date knowable | Accept | Explicit deployment identifiers; legacy and new BOLD series separated; January 23 supported by contemporaneous DeFi Saver and legacy history, May 19 by Liquity. i1, c19, c21, s1. |
| A-E4 | Audit changes are knowable; exact February root cause remains unverified | Accept | Read ChainSecurity §2.2.13 and diffed January/May SP files. Documented 1-BOLD residual, product precision and epoch/error changes without equating them to a proven exploit path. i2, c20. |
| A-E5 | Report/data mapping, exact dates and labels inconsistent | Accept | Every labeled report block now maps to one of 68 dataset claims. Dates explicitly mean read dates, with event/block dates in text/figures. Recommendations have their own inference claims. All original c1–c27 IDs remain. |
| A-final | Original unsuitable for choosing launch parameters | Accept | Revised conclusions still do not certify optimal parameters or a hook. Added practical evidence and a clear separation between reference constants, observations and proposed tests. §7; G9. |

## Critique B: mechanism and supposedly fillable gaps

| Point | Critique | Decision | Independent source check and change |
|---|---|---|---|
| B0 | Core questions lack observations; fork incidents/reflexive risk omitted | Accept | Added empirical panels and USDaf/Nerite evidence; introduced Synthetix as a distinct reflexive comparator. Limits remain explicit rather than claiming complete coverage. §§2–8. |
| B1.1 | Allocation is outside debt, not a merged rate queue | Accept | Verified directly in contracts, corrected c15, added measured incidence and frozen-state allocation example. |
| B1.2 | SP coverage determines branch exposure; thin pool always receives more | Partly accept | Outside debt is **absolute debt minus SP**, floored at zero; coverage ratio alone does not rank branches. Rate levels do not directly set allocation. Clarified d3; a small thin-SP branch need not take more absolute volume than a large branch. |
| B1.3 | Branch frontiers differ; well-covered branch permits lower rates | Partly accept | Different frontiers follow from separate lists. No universal assurance of safety: all-zero-outside-debt fallback and later SP changes matter. Snapshot rETH coverage/high rates illustrate heterogeneity. c15, r4, b1. |
| B1.4 | Redeemers can express preference through direct ordinary TroveManager calls | Reject | `redeemCollateral` calls `_requireCallerIsCollateralRegistry`; direct external normal redemption is prohibited in reviewed code. Urgent shutdown redemption is different. c15, i3. |
| B1.5 | Check fork routing and focus monitoring on SP allocation | Partly accept | Nerite, Orki and Felix docs checked; USDaf incident checked; Quill legacy/current identity flagged. Full deployed-code comparison for all forks remains G8. Monitoring recommendation corrected to outside debt, with temporary-deposit manipulation also documented. i5, d3. |
| B2.1 | Fee floor/initial rate/decay/beta/urgent bonus exist in source | Accept | Checked May audited constants against deployed registry's verified source bundle; files identical. Added 0.5%, 100%, six hours, beta 1 and 2% respectively, with launch-comment caveat. c2, c14. |
| B2.2 | Seven-day duration/cooldown is not uncertain; fee not on every adjustment | Accept | Verified BorrowerOperations and constants; revised f1. c2 label now consistent. |
| B2.3 | Q6 should include MIN_DEBT, rates, MCR/CCR and SP split | Accept | Added reference table, preserving distinction between per-trove minimum and minimum viable supply. Deployed source bundle qualification stated. c2, c27. |
| B2.4 | oordinal dashboard, Dune or logs can fill redemption gaps | Accept | Retrieved pinned export and pricing/collection code, reconciled hit totals, spot-checked six events in one receipt; added monthly/wave/branch figures. Dune shell supplied no data. m1, c12–e5, b1. |
| B2.5 | “No official root cause exists”; P-rounding issue may identify it | Partly accept | SP precision diff and audit changes confirmed. Secured-Finance issue explicitly cites a v1 advisory, so it cannot identify February BOLD root cause. Neither universal absence of a disclosure nor exact exploit identity is established. i2, c20. |
| B2.6 | Frontend understated debt-in-front outside 0.1% brackets | Accept | Full v1.8.1 release body confirms the defect and pending-debt omission; v1.5.0 also corrected per-branch calculation. Added dated measurement caveat. c25, gap1. |

## Critique B: forks, incidents and attribution

| Point | Critique | Decision | Independent source check and change |
|---|---|---|---|
| B3.1 | Fork survey must identify Felix, Quill, Orki, USDaf, Nerite, Ebisu/chains/dates | Accept | Official dated friendly-fork announcements checked. Added chains, dates, separate deployment supply histories, and explicit Ebisu gap. Felix date identified as public staged-launch milestone. c23, s1, i5. |
| B3.2 | Missed June 19/20 USDaf oracle shutdown | Accept | Read full June 22 team post-mortem. Added threshold/equality mismatch, fallback failure, other branches without fallback, unwind and attributed no-loss claim. June 19 incident distinguished from June 20 notice. c22. |
| B3.3 | Describe shutdown/urgent redemption and Coinspect BOLD-05 | Partly accept | Urgent path verified in code and added. BOLD-05 finding/resolution was not independently extracted; retained as unverified lead, not a confirmed cross-branch failure. i3, i8. |
| B3.4 | Supply survivorship missing; July 2026 USDaf ~$289k/~95% decline lead | Partly accept | Replaced unverified secondary statistic with directly retrieved dated DefiLlama histories for both USDaf deployments and other forks. Large contractions documented; causal size/peg claim not inferred. s1. |
| B3.5 | Add Prisma and Raft; Raft bug same class as suspected BOLD | Partly accept | Added Prisma team/response post-mortem with reported loss and response. Raft primary report not recovered; not counted as v2 incident. No claim that Raft identifies BOLD's bug. i7–i8, c20. |
| B3.6 | October 10, 2025 crash omitted | Accept | Opened full DefiLlama adapter PR body: reported liquidation outputs confirmed as author test results. Added BOLD daily prices, no-redemption observation and historical mean-rate snapshots. No unsupported peg/SP intraday narrative. i6, r1–r2. |
| B4.1 | V1 account too reliant on Liquity; Parsec body not read | Accept | Added 2,193 direct LUSD logs, same-year totals and sampled prices. Retrieved Parsec body. DSR explanation remains attributed to Liquity. c7–c11, p1. |
| B4.2 | SP “unaffected” scope is self-report, migration/reward costs omitted | Accept | Qualified assurances; preserved exit/migration/reward cutoff facts without invented motive. Added legacy supply/price stress; did not infer every user paid a particular amount or lost funds. c16–c21. |
| B4.3 | Design intent must not be empirical evidence | Accept | c3/c10 explicitly describe source intent. Rate observations, event counts and recommendations are now separate claims with consistent labels. |
| B5.1 | No post-May-2025 figures; FAQ weekly example indicates activity | Accept, with caveat | Added 2026 panels. FAQ's 200k-versus-15m weekly numbers are illustrative, not a dated measured wave; they are not used as observations. c12, e2, r1, s1. |
| B5.2 | Initial launch date need not be uncertain | Accept | January 23 verified through contemporaneous integration announcement and legacy series. May 19 distinctly relaunched deployment. i1, c19. |
| B5.3 | LUSD comparison stops at 2023 despite coexistence | Accept | LUSD logs cover all four years through October cutoff; added annual volumes/fees and selected same-day comparisons. c11, e3. |

## Critique B: recommendations and disputed replacement claims

| Point | Critique | Decision | Independent source check and change |
|---|---|---|---|
| B6.1a | Reflexive staked-IMD collateral is missing | Accept | Added endogenous confidence/exit-liquidity stress analysis; ETH/LST observations cannot calibrate IMD. d5. |
| B6.1b | Synthetix 2025–2026 should provide a comparator | Partly accept | Checked SIP-420, official May 15 peg update and SIP-423; used $0.93 contemporaneous team report and June 2026 retirement specification. Unverified $0.66/$0.21 lows not repeated; SIP status not treated as proof every phase completed. d4. |
| B6.1c | Test IMD 50%–80% falls and revisit prior swarm | Partly accept | Added 50%/80% as hypothetical test severities, not observed or calibrated parameters. Prior campaign was not read, so no fabricated citation or re-covering 2020–2021. d5, gap3. |
| B6.2a | External-yield rate-floor policy unproven; test sudden yield shock | Accept | Withdrew policy, added proposed competing-yield/participation stress test and implementation floor. d2. |
| B6.2b | Seven-day cooldown prevents faster borrower reaction | Reject as a general rule | Ordinary premature changes cost a fee; they are not universally locked for seven days. Safety Mode may prohibit fee-incurring changes. f1, g4. |
| B6.2c | 0.5% floor plus 100% initial base rate is Liquity's launch answer | Partly accept | Constants confirmed as implementation choices, not proof of launch optimality or that initial base rate stayed high. c2, c14, d1–d2. |
| B6.3a | No SP/debt or executable depth basis for minimum-size gate | Accept | Added measured SP/debt snapshot and supply trajectories. No executable-depth calibration exists, so numerical gate remains unsupported. b1, c27. |
| B6.3b | MIN_DEBT and zombies missing; all zombies always first | Partly accept | Added actual loop and `lastZombieTroveId` exception. Only the pending nonzero zombie gets that treatment; categorical all-zombie claim rejected. g2. |
| B6.3c | Splitting has structural consequences | Accept | Withdrew blanket dismissal; discuss queue ties, minimums, zombie transition, fee/gas and different-rate slices without claiming observed profitable evasion. g3. |
| B6.4a | Report overlooks upper-side minting/soft ceiling | Accept | Narrowed conclusion to lack of a tight upper defence from redemption itself; added mint/sell incentives and v1 MCR-related boundary. c26. |
| B6.4b | Generic mint ceiling is `1/(1−fee)` | Reject as a guarantee | Formula omits collateral commitment, liquidation/financing/exit conditions. V1 documentation's broad MCR argument is not that formula and is not a tight $1 defence. c26. |
| B6.4c | Hook response has no inventory analysis; dynamic fee/oracle manipulation omitted | Accept | Removed validated-remedy language; added explicit test proposals, including the defended pool setting collateral borrowing power. No unseen hook certified. d6. |
| B6.5 | Oracle failure/shutdown must be modelled for new collateral | Accept | USDaf report and urgent path checked, added staleness/equality failure and irreversible-unwind scenario to design implications. c22, i3, d3, d5. |
| B6.6a | MEV discussion theoretical; Electisec front-running finding lead | Partly accept | Builder/priority/base-fee estimates now quantified; alleged audit finding remains uncertain because audit page returned only shell. Costs do not prove borrower front-running. g5, c24. |
| B6.6b | Oracle-gap arbitrage can redeem near $1, not just during underpeg | Accept | Source code plus export confirm 259 above/parity-quote transactions totaling 8.203m BOLD. Added two 2026 days and explicit executable-price caveat. e4–e5. |
| B7.1 | Survivorship bias includes too little fork adversity/success | Accept | Added shrinking fork series, USDaf unwind, Nerite's current Kelp-exposure warning and October 2025 liquidation comparison. No claim that shrinking supply equals failed peg or that no further incidents exist. s1, c22, i4, i6. |
| B7.2 | Name delegates, fees and batch entry/exit rules | Accept, with caveat | Official metadata identifies managers; snapshot quantifies fees/adoption. Joining incurs upfront fee; do not generalize all entry/exit behavior into a universal one-week prohibition. c4, r5, f1. |
| B7.3 | c2/c9/c15/c26 labels conflict or false premise persists | Accept | c2 duration verified; c9 body retrieved; c15 false premise removed; c26 consistently inference with narrower proposition. One-to-one report/dataset mapping checked. |

## Critique B: revision priorities, tracked separately

| Point | Requested priority | Decision | Result |
|---|---|---|---|
| B8.1 | Correct cross-branch rule | Accept | c15, b1 and d3 corrected and quantified; direct normal bypass claim rejected. |
| B8.2 | Add reference constants tied to deployment | Accept | c2/c14; verified registry source bundle matches audited May constants exactly. Remaining branch-by-branch bytecode-equivalence limit stated. |
| B8.3 | Pull post-relaunch event/price/rate data and fork supplies | Accept | r1–r5, c11–e5, b1, s1; full panels and causal claims still marked gaps. |
| B8.4 | Add USDaf, shutdown, urgent redemption and zombies | Accept | c22, i3, g2. |
| B8.5 | Add reflexive comparator and retest implications/oracle stress | Accept | d1–d6 separate implemented values from unsupported optimum; Synthetix is explicitly not a Liquity fork. |
| B8.6 | Relabel self-reports and fix c2/c9/c26 | Accept | c8, c16, c19, c22 retain attribution; c2 verified, c9 directly read, c26 narrowed. |

## Checks and limits

**[fact]** The accompanying offline check validates evidence-file hashes, JSON shape/claim IDs, source URL shape, reconciliation of BOLD transactions with rounded hit amounts, the selected independent receipt, historical tuple shapes and weighted means, LUSD log deduplication/decoding, annual/monthly totals, wave observations and report/dataset labels. It does not independently certify the dashboard's full indexing, price execution, every audit finding, or causal design recommendations.

**[uncertain]** The task's broad question is better answered but not exhausted: a full manager/manual exposure panel, complete fork on-chain histories and intraday executable peg paths remain outstanding. Archive reads first failed on PublicNode because they required a personal token; dRPC successfully supplied the historical reads. Etherscan returned 403, Dune a page shell, and the Electisec page only a shell; Blockscout and public exports supplied the usable evidence. No requester-only inputs were needed.