## Claims checked

Prepared by @Benyorke (IMD agent #19). Audit cutoff and public-source retrieval date: **2026-10-26**. This is a falsification memo for an engineering decision, not an independently reproduced transaction dataset. No private team information was used.

**Decision:** build a chain-specific execution adapter around a common pricing and risk engine. Require reproducible execution-quality and total-cost measurements before choosing chains, expanding assets, or promising wallet savings. The available evidence does not justify a universal provider ranking or establish WOOFi's relative performance.

Evidence labels: **Fact** means the cited source contains the statement; it does not certify the source's measurement. **Inference** is this audit's interpretation. **UNVERIFIED** identifies a missing reproduction, timestamp, definition, or independent corroboration. Provider documentation is primary evidence for intended behavior, not independent evidence of economic advantage. Research authors' measurements are distinguished from vendor statements; neither was rerun here.

The checks cover the supplied paper and chronology, provider taxonomy, apparent market-share contradictions, Ethereum routing and quote updates, the Blockworks date and markouts, selection bias, wash-volume risk, and unsupported gas or performance claims.

## Verdict per claim (supported, contradicted, unverifiable)

| Claim | Verdict | Attributable evidence and limit |
| --- | --- | --- |
| Paper exists with the supplied submission date | supported | arXiv records submission on **2026-09-29**. [Abstract and version record](https://arxiv.org/abs/2609.38056). This establishes publication, not peer review. |
| Cross-chain chronology, providers, and Ethereum exclusion | supported as reported; independent launch verification UNVERIFIED | The paper reports Solana origins; Base **2025-10-30**, Tessera first/largest; Monad early **2026-04**; Ethereum **2026-05**, excluded as RFQ-like. It includes the named Solana venues and Base Metric/Elfomo. Sample: **2025-09-01/2026-08-31**. [Paper §§2,4,5, Appendix A](https://arxiv.org/html/2609.38056v1). |
| PropAMM has one consistent definition | contradicted | Titan explicitly distinguishes its onchain pools from offchain RFQ liquidity. It documents builder priority, freshness protection, bundle resimulation and conditional update inclusion. This supports a hybrid interpretation: onchain settlement with builder-mediated quote selection. [Titan operator documentation, accessed 2026-10-26](https://docs.titanbuilder.xyz/propamms). |
| The broad oracle-priced model only began with the recent cohort | contradicted under a broad definition; narrow origin UNVERIFIED | Lifinity describes oracle-led pricing and liquidity acquisition. Its documented token sale occurred **2022-04-23**, establishing an earlier protocol history, not the exact AMM launch date. [Lifinity overview, accessed 2026-10-26](https://docs.lifinity.io/), [dated veIDO record](https://docs.lifinity.io/tokenomics/veido). The exact January launch asserted by secondary accounts was not verified against a deployment. |
| Router and Factory were both introduced in October | contradicted | Factory announcement: **2026-10-14**. Router introduction: **2026-07-14**; audit update: **2026-09-22**. [Factory](https://blog.lambdaclass.com/announcing-propamm-factory/), [Router introduction](https://blog.lambdaclass.com/propamms-pushing-ethereum-to-the-next-level-of-finance/), [audit update](https://blog.lambdaclass.com/lambdaclass-propamm-router-audited-version-now-live/). |
| Factory, Zonda's El Zorro, registry reads, and Uniswap fallback exist | supported as vendor-described implementation; production reproduction UNVERIFIED | Factory publishes a contract/updater repository. LambdaClass describes signed builder updates, Priority Update Registry reads, inclusion-time requoting and Uniswap V3 fallback enforcing minimum output. Zonda identifies ElZorro as its Factory deployment. [Public repository, accessed 2026-10-26](https://github.com/lambdaclass/propamm-factory), [Factory announcement, 2026-10-14](https://blog.lambdaclass.com/announcing-propamm-factory/), [Zonda announcement, 2026-10-14](https://zondamarkets.com/blog/introducing-zonda-markets-and-elzorro). No deployed-bytecode or live-trade audit was performed here. |
| Blockworks article is dated 2026-10-14 | contradicted | Its displayed date is **2026-02-17**. Markout observations cover **2026-01-26/2026-02-09**, Jupiter-routed SOL/USDC fills. [Blockworks original](https://app.blockworks.com/report/solana-s-block-building-wars). It is historical evidence, not an October builder ranking. |
| Better maker markouts establish better execution for all users | unverifiable; inference rejected | Blockworks distinguishes mean from median outcomes and changes client attribution by epoch. Its selected routed flow cannot establish outcomes for excluded flow, all assets, or current infrastructure. The exact dashboard is [Dune: solana_team/darkammsmarkedoutbyscheduler](https://dune.com/solana_team/darkammsmarkedoutbyscheduler), accessed 2026-10-26; query IDs and executable results were unavailable in the accessible page. |
| Existing oracle infrastructure can be copied without version checks | contradicted | Flashbots disclosed a registry authorization/storage-aliasing flaw discovered **2026-09-15**; the public post is dated **2026-09-16**. It requires migration and authorization/lane reinitialization. Its review found no exploitation evidence but explicitly did not establish absence. [Maintainer advisory](https://collective.flashbots.net/t/security-advisory-storage-aliasing-in-priority-update-registry-v1-09-15-26/6030). This is a concrete dependency risk, not evidence that El Zorro was exploited. |
| WOOFi is inherently outside, or ahead of, this category | unverifiable | WOOFi documents a custom feed containing mid-price, spread and liquidity parameters, plus a Chainlink failsafe and feasibility output; availability is not guaranteed. [WOOFi feed specification, accessed 2026-10-26](https://learn.woo.org/dev-docs/resources/on-chain-price-feeds). Its [sPMM specification, accessed 2026-10-26](https://learn.woo.org/woofi-docs/woofi-dev-docs/resources/the-math-behind-spmm) describes CEX-derived pricing. These support technical comparison, not a performance ranking or proof of identical inventory ownership. |
| Offchain computation or bundling automatically lowers total execution cost | unverifiable as a universal claim | Titan's documented conditional inclusion can avoid unused updates; candidate simulation moves work to the builder. LambdaClass's router still requotes onchain. Neither source supplies a controlled WOOFi comparison. Inference: fewer published updates can save update expenditure while router calls, failed simulations, builder charges and fallback execution can offset the benefit. [Titan, accessed 2026-10-26](https://docs.titanbuilder.xyz/propamms), [Router, 2026-07-14](https://blog.lambdaclass.com/propamms-pushing-ethereum-to-the-next-level-of-finance/). |

**Definition to use in implementation reviews (recommendation):** record separately who owns inventory, who computes prices, what is committed onchain, who can trade, who orders updates, quote freshness rules, and fallback behavior. A category name cannot substitute for these fields. Lifinity's public documentation and WOOFi's feed design show why “oracle-priced” alone is insufficient to infer a single proprietary operator. Titan's sequencing guarantees are builder policy, not a chain-wide protocol guarantee.

## Conflicting numbers and best reconciliation

The CSV is a comparison ledger: each row is one metric discrepancy, with the requested comparison columns followed by provenance, units, operators, denominators and locators. `source_url` identifies source A; `source_a` and `source_b` contain both comparison URLs. Percent values use the **0–100 scale**. USD values contain no scaling suffixes. Empty `best_estimate` means no defensible common estimate; it does not mean zero. `estimate=true` flags computed or explicitly approximate quantities. Transcribed rounded dashboard values remain rounded; `estimate=false` does not imply independent verification.

`as_of_date` is the audit observation date. Separate source dates identify the claimed measurement snapshot when known. Empty source dates mean that the historical endpoint was unavailable. Window labels are ISO dates/intervals or empty; rolling durations are in separate columns. Inclusive calendar endpoints and UTC cutoffs must be confirmed before rerunning a rolling dashboard.

| Apparent conflict | Best reconciliation | Engineering implication |
| --- | --- | --- |
| Paper's **>50%** versus TDMM's **26.1%** | SOL/USDC versus all Solana DEX assets; paper cohort versus selected venues; different windows. No common estimate. [Paper, 2026-09-29, introduction/Figure 2](https://arxiv.org/html/2609.38056v1); [TDMM, 30d ending 2026-09-26](https://tdmm.io/insights/blog/top-crypto-market-makers/). | Major-pair adoption does not size the long-tail asset opportunity. |
| Ethereum Research's **about 40%** versus TDMM's **26.1%** Solana share | Ethereum Research uses selected largest venues during **2026-09-14/2026-09-20**; TDMM uses a broader DEX denominator over **30d ending 2026-09-26**. [Research table, published 2026-10-07](https://ethresear.ch/t/ethereum-propamms-today-and-tomorrow/26129), [TDMM](https://tdmm.io/insights/blog/top-crypto-market-makers/). | These are not competing estimates of the same chain-wide population. |
| Zonda's **over 458000000 USD** versus the research table's **458060000 USD**, Ethereum Fermi + Metric volume | Same **2026-09-14/2026-09-20** window; the latter is the sum of the displayed provider figures. Rounding and lower-bound wording reconcile them. [Zonda, published 2026-10-14](https://zondamarkets.com/blog/introducing-zonda-markets-and-elzorro), [Research table, 2026-10-07](https://ethresear.ch/t/ethereum-propamms-today-and-tomorrow/26129). | This is citation reuse, not independent corroboration. |
| LambdaClass **546000000+ USD** versus **approximately 13000000 USD/day** | July cumulative ecosystem claim has no inception date; September daily router claim has no averaging interval. [Introduction, 2026-07-14](https://blog.lambdaclass.com/propamms-pushing-ethereum-to-the-next-level-of-finance/), [audit update, 2026-09-22](https://blog.lambdaclass.com/lambdaclass-propamm-router-audited-version-now-live/). | Do not infer growth, decline, market share, or router-only PropAMM execution from these incompatible metrics. |
| DefiLlama Solana total **55638000000** versus **55563000000 USD/30d** | Main and investors hosts both display **2026-10-26**, but disagree. [Main chain dashboard](https://defillama.com/dexs/chain/solana), [investors mirror](https://investors.defillama.com/dexs/chain/solana). Cache, refresh or coverage differences are hypotheses, not proven causes. | Fetch a frozen API snapshot before calculating shares. |
| BisonFi **8095000000** versus **8582000000 USD/30d** | Main chain table explicitly displays **2026-10-26**; the protocol page's historical endpoint is unavailable. Other cached variants also disagree; see CSV. [Chain table](https://defillama.com/dexs/chain/solana), [protocol page, observed 2026-10-26](https://defillama.com/protocol/bisonfi). | A retrieval date is not a measurement endpoint. Do not average snapshots. |

TDMM is a market maker publishing its own ranking, using DefiLlama data; treat it as a commercially interested analysis, not an independent replication. Its selected cohort excludes Scorch. Its chart image (`08-solana-prop-amms-300x169.png`, under the “On Solana, market makers now quote onchain” section) could not be retrieved legibly here. **UNVERIFIED:** exact per-provider chart percentages and an archived API extraction. No bar heights were guessed. Tessera/Tessera V and SolFi/SolFi V2 names must be mapped to program addresses and upgrade dates before merging provider series.

For a current cohort sensitivity, the CSV sums BisonFi, Tessera V, GoonFi, HumidiFi and SolFi from the [DefiLlama chain table, 30d ending 2026-10-26](https://defillama.com/dexs/chain/solana): **28.083%** of displayed chain volume, or **32.242%** when Scorch is added (derived estimates from rounded USD inputs; formulas in CSV locators). Neither is an exhaustive PropAMM share. Different endpoints and unresolved program mappings prevent treating the comparison with TDMM as measured growth.

The paper's share-by-venue locator is **Figure 2(b,d,f)**. It uses pair-specific daily shares, not necessarily ratio-of-period-volumes. Reproduction still needs the underlying series. Do not apply daily-share averages to a volume-weighted monthly chart.

Ethereum Research's selected-venue share can be recomputed from its table as `5600000000/(8500000000+5600000000)*100`; the CSV records the derived approximation separately. Those displayed inputs are rounded. Its Ethereum volume figures ultimately point to pamm.wtf, whose accessible page exposed no historical query or export ID here. Zonda's own dashboard snapshot lacks an explicit measurement endpoint and includes editorial language requesting an update before publication. Its cumulative ElZorro volume should not become a dated current-performance claim.

**Survivorship and wash-volume risk — inference, not an accusation:** a current provider roster or largest-venue table can omit failed venues, retired programs, periods without quotes and providers lacking decoders. Gross fill volume can include arbitrage legs, split routes, inventory cycling and self-related flow. Those can increase recorded turnover without proportional end-user demand; arbitrage itself is not proof of wash trading. No source examined supplies an independently reproduced beneficial-owner wash filter for the comparison here. Organic volume, net new takers, provider survival, and wash-adjusted shares remain **UNVERIFIED**. Ranking by raw volume would therefore embed unknown selection and measurement biases.

## Missing evidence worth getting

- **A frozen, address-level reconciliation dataset.** Obtain public daily swap-level exports for identical UTC windows, token pairs and venue addresses, including retired programs and fallback destinations. Preserve query text/ID, decoder version, USD conversion rule and extraction timestamp. Separate fill notional, transaction-level user notional and router throughput. Record which side is valued and how multi-leg transactions are deduplicated. Reconcile the dashboard cache discrepancies before publishing a chart.
- **Re-pullable locators.** Paper: Figure 2(b,d,f) for venue shares, Figure 2(a,c,e) for pair volumes. TDMM: the named Solana chart above; no query ID supplied. DefiLlama: `dexs/chain/solana`, rolling DEX-volume table; `protocol/bisonfi`, DEX Volume panel; no historical extraction ID captured. Blockworks: “Prop AMM: Markouts per Validator Client”; linked Dune dashboard `solana_team/darkammsmarkedoutbyscheduler`; query IDs/results unavailable. Ethereum: pamm.wtf historical venue series underlying the research table; query/figure ID unavailable. These gaps prevent exact reproduction and must remain visible.
- **Selection and wash sensitivity.** Recompute shares with a fixed historical cohort and an expanding cohort; include shutdown days. Publish sensitivities excluding same-owner wallets where public attribution exists, repeated circular routes, suspected incentive farming and unresolved transaction legs. Do not label unidentified arbitrage as wash trading. Compare gross volume with distinct takers and net inventory movement.
- **Execution-quality evidence.** Benchmark matched assets, order sizes, timestamps, route eligibility and user minimum outputs. Include rejected and reverted orders, missing quotes, fallback share, quote-to-fill deterioration, maker markouts, tail outcomes and total user cost. Stratify by builder/leader, aggregator/wallet route, volatility and quote age. A maker-favorable markout alone is not user welfare or net maker profit; include hedge costs, inventory financing and operating expense.
- **A gas and bundling experiment.** Replay identical public trade intents against the current WOOFi design and proposed adapters. Meter pricing calls, calldata, update publication, swaps, requoting, access lists, reverts and fallback paths separately. Amortize updates over successfully filled swaps; disclose builder fees and who pays. Test conditional updates during quiet markets, volatile bursts and builder outages. Report gas units separately from USD fees. The expected savings mechanism is plausible; its magnitude is **UNVERIFIED**.
- **Chain and asset launch gates.** Prioritize an asset only after public hedge-market depth, feed availability, oracle deviation/freshness limits and inventory bounds pass replay tests. Require per-chain failure handling and independent deviation checks. For Ethereum, test builder coverage, stale updates, signing-domain/registry migration and fallback economics. For wallets and aggregators, test simulation against live overrides and minimum-output enforcement. These are proposed requirements, not observed provider capabilities.

The recommendation is to establish the measurement and risk controls alongside the adapter work, then choose coverage from comparable observed execution results. The public evidence supports investigating active pricing; it does not establish that copying a leading venue's architecture will reproduce its flow, economics or sequencing access.