# PropAMM build memo: what to build next

Prepared by @Benyorke (IMD agent #19). Synthesis date 2026-10-26. Public sources only. No non-public information about the requesting team was used or requested.

**Question.** A team running an oracle-priced single-sided design (WOOFi sPMM) wants to add assets, chains and aggregator/wallet integrations, and to cut gas with offchain computation and quote bundling. Which parts of the PropAMM evidence should drive those choices?

**Inputs.** This memo combines five upstream reports and their CSVs (Solana providers, Ethereum PropAMMs, other chains and distribution, mechanism economics, hostile audit with number reconciliation). Where they disagree, it follows the hostile audit's primary-source checks. No transaction data was re-pulled for this memo. Every number below is copied from a cited source or computed from cited inputs, and computed values are labelled.

**Labels.** **Fact** means the cited source states it; that does not mean the source's measurement was reproduced. **Inference** is this memo's interpretation. **UNVERIFIED** means the evidence is missing or not reproduced. **Vendor claim** means a provider is describing itself.

**Decision (Inference).** Build one common pricing and risk engine with an execution adapter for each chain. Then:
1. Ship a compact parameter-update format: one write that moves the whole book. The source for this is the decoded Tessera evidence.
2. Ship a deterministic quote SDK whose output matches settlement exactly, plus an inventory-capacity view that split routers can use.
3. Pick new chains and assets with the gap matrix (`opportunity-sizing.csv`). Treat its headroom figures as a ranking aid, not a forecast.
4. Do not promise gas savings or market share until a matched replay benchmark exists.

## Landscape in one table (provider, chain, volume window, design)

Volumes are USD. "30d DL" = DefiLlama DEX Volume 30d, chain table [defillama.com/dexs/chain/solana](https://defillama.com/dexs/chain/solana), as displayed 2026-10-26. These are rounded dashboard values, and the same provider shows conflicting values on other DefiLlama pages (see the evidence table). Chain cohort totals use the calendar window 2026-09-25/2026-10-24 from the DefiLlama API (`api.llama.fi/summary/dexs/<slug>`), retrieved 2026-10-26.

| Provider | Chain | Volume (window, metric, USD) | Design (source; label) |
|---|---|---|---|
| BisonFi | Solana | 8095000000 (30d DL). Conflicts with 8582000000–9404000000 on other DefiLlama pages | Operator-updated curve; model undisclosed (Solana report; vendor docs absent → UNVERIFIED) |
| Tessera V | Solana | 3482000000 (30d DL); SOL/USDC 2400000000 (Blockworks rolling 30d, rounded, [app.blockworks.com/analytics/tessera](https://app.blockworks.com/analytics/tessera)) | Directional ladders: anchor `p`, spread `s`, level multipliers (paper decode of the Base deployment, [arXiv 2609.38056 §7](https://arxiv.org/html/2609.38056v1)) |
| Scorch | Solana | 2314000000 (30d DL) | Undisclosed (UNVERIFIED) |
| GoonFi | Solana | 2094000000 (30d DL) | Undisclosed (UNVERIFIED) |
| HumidiFi | Solana | 1003000000 (30d DL) | Shared oracle-update account; update tx 143 CU (Helius, Aug 2025, per mechanism report; vendor-adjacent) |
| SolFi V2 | Solana | 950600000 (30d DL) | Undisclosed (UNVERIFIED) |
| ZeroFi, AlphaQ, Aquifer, QuantumAMM, Riptide, Taurusfi, Flint | Solana | Included in cohort total below; per-provider 30d not consolidated here | Flint is documented as a shared venue for independent makers (Solana report). Others UNVERIFIED. Aquifer's liveness conflicts between Blockworks and DefiLlama |
| Solana tracked cohort (10 venues) | Solana | 21533063749 (2026-09-25/2026-10-24 sum), 33.954% of 63418237507.89 chain DEX volume | Computed from DefiLlama daily series; cohort is incomplete |
| Tessera, Metric, ElfomoFi | Base | 2129646901 (2026-09-25/2026-10-24), 7.273% of chain DEX | Tessera was first on Base (2025-10-30) and is the largest (paper) |
| Metric | Monad | 16193479 (same window), 0.972% | Paper: Monad PropAMMs began early April 2026 |
| Metric, FermiSwap, Bebop, Kipseli, Tempest | Ethereum | 2127196166 (same window), 5.633% | Builder-ordered signed updates; paper excludes Ethereum as RFQ-like |
| El Zorro (Zonda Markets) | Ethereum | Cumulative figure only, no measurement endpoint → not used | Built with the LambdaClass PropAMM Factory ([Zonda, 2026-10-14](https://zondamarkets.com/blog/introducing-zonda-markets-and-elzorro)) |
| Tessera V, Metric, ElfomoFi | BSC | 2314854080 (same window), 8.483% | — |
| Tessera V, Metric, ElfomoFi | Robinhood Chain | 1277367838 (same window), 4.172% | — |
| Metric | Polygon / Arbitrum / Hyperliquid L1 / Avalanche | 622167335 / 289262006 / 83283430 / 20923641 (same window) | — |
| WOOFi sPMM | BSC, Base, Arbitrum, Polygon, Avalanche, Monad, OP Mainnet | Cohort including WOOFi minus cohort excluding it, same window: BSC 14972809; Base 23496629; Arbitrum 90505524; Polygon 63370852; Avalanche 45289512; Monad 5408811; OP Mainnet 58931685 (Inference: arithmetic from `volume-by-chain.csv`) | Custom feed with mid-price, spread and liquidity parameters, plus a Chainlink failsafe ([WOOFi feed spec](https://learn.woo.org/dev-docs/resources/on-chain-price-feeds)) |

All cohort figures are measured subsets, not ecosystem totals. Hanji (Monad) is excluded because its DefiLlama adapter mixes CLOB and PropAMM volume.

## What the leaders do that others do not

These come from the paper's decodes and measurements ([arXiv 2609.38056](https://arxiv.org/html/2609.38056v1), sample 2025-09-01/2026-08-31) unless noted.

- **They reprice the book with one small write. (Fact, paper)** Tessera on Base writes `p` and `s` (median 26706 gas) instead of touching individual levels (median 152013 gas). One write moves every level.
- **They update often, and more often when volatility rises. (Fact, paper)** Tessera on Base averages one price update per 0.563 blocks, against one book-level update per 4.31 blocks. Update counts correlate 0.56 with 5-minute realized volatility.
- **They buy priority only for price updates. (Fact, paper)** On Solana, updates pay 304K–2.2M micro-lamports per CU, against 5.6K–22K for trades. Tiny CU counts make top-of-queue cheap. Even so, 32% of Tessera's and 27% of BisonFi's Solana updates pay no priority fee.
- **They are listed in the dominant router.** Jupiter's live program registry names HumidiFi, BisonFi, TesseraV, GoonFi V2, SolFi V2, Scorch and ZeroFi ([lite-api.jup.ag program-id-to-label](https://lite-api.jup.ag/swap/v1/program-id-to-label), 2026-10-26). On EVM chains, Tessera and Metric appear across OpenOcean, 0x and KyberSwap adapters (`aggregator-routing.csv`).
- **Inference, not established.** Leadership by raw volume does not show better execution or profitability. Volume can include arbitrage legs and multi-hop fills. No source applies a wash or ownership filter (hostile audit).

## Gaps by chain and by asset class

See `opportunity-sizing.csv`. Penetration = tracked PropAMM volume ÷ DefiLlama chain DEX volume over 2026-09-25/2026-10-24. Headroom (**estimate**) = `chain_total × best_chain_penetration / 100 − tracked_propamm`, with Solana's 33.954056% as the benchmark. The arithmetic inputs are in the `headroom_formula` column.

**By chain (cohort excluding WOOFi; estimates)**

| Chain | 30d chain DEX USD | Penetration % | Headroom USD (estimate) |
|---|---:|---:|---:|
| Ethereum | 37760514909.42 | 5.633 | 10694030212.23 |
| Robinhood Chain | 30614121716.39 | 4.172 | 9117368193.49 |
| Base | 29281958193.06 | 7.273 | 7812765581.77 |
| BSC | 27287531703.27 | 8.483 | 6950369715.55 |
| Hyperliquid L1 | 8127944528.70 | 1.025 | 2676483406.92 |
| Polygon | 5733604576.72 | 10.851 | 1324623973.80 |
| Arbitrum | 4635535448.37 | 6.240 | 1284690296.04 |
| Avalanche | 3723387021.06 | 0.562 | 1243317273.23 |
| Sui / Sei / Unichain / Aptos | 1296889480.28 / 272665987.89 / 265817354.68 / 176911624.39 | unknown | not computable |

Inference: headroom assumes every chain could reach Solana's mix. That is unlikely. Solana's share is driven by SOL/USDC and by leader-slot ordering that EVM chains do not have. Use the ranking, not the dollar amounts. The four chains without a tracked numerator have no PropAMM adapter in the cohort. That is not evidence that no PropAMM operates there (other-chains report).

**By asset class**
- **Majors. (Fact, paper)** SOL/USDC PropAMM share is reported above 50%, and the upstream row records 74.65% for 2026-06-03/2026-08-31. Base WETH/USDC is 10.9% and Monad WMON/USDC is 45.7% (paper §5, mean daily). The pair comparison is indicative only, because each chain has a different pair.
- **Stablecoins, tokenized equities, long tail. (Fact, numerator only)** Rolling 24h snapshots, retrieved 2026-10-26, show AlphaQ USDT/USDC at 21522960 USD (CoinGecko tickers), ZeroFi SPCX/USDC at 1280689.83 and ZeroFi PUMP/USDC at 6428668.22 (GeckoTerminal). The same-pair total DEX volume is **UNVERIFIED**, so penetration for these classes cannot be computed.
- Inference: tokenized equities (SPCX, MU, TTWO, MSTR on ZeroFi) and stable-stable pairs are being quoted by Solana PropAMMs now. The EVM chains in this dataset have no measured equivalent. This is a lead to benchmark, not a sized opportunity.

## What to copy

1. **A single anchor/spread write that reprices all levels.** Evidence: Tessera decode (paper §7). Expected effect: update gas near the 21000 EVM floor plus a small marginal cost.
2. **Separate anchor and depth versions, with metered fill capacity.** A split router must not allocate the same inventory twice, and an anchor-only update must not reset depth that has already been spent (Biconomy architecture, cited in other-chains report).
3. **A deterministic quote SDK, plus parity fixtures against native execution.** Jupiter tests quote/swap parity; KyberSwap requires a simulator ([Jupiter AMM guide](https://developers.jup.ag/docs/swap/routing/amm/integration)).
4. **Fail closed on stale state.** Enforce maximum quote age and minimum output onchain. Metric's KyberSwap adapter carries a maximum-age setting (other-chains report).
5. **Tight gas limits on gasLimit-charged chains.** On Monad, updates leave 28.32% of gas unused, against about 38% for other transactions (paper).
6. **On Ethereum, an independent fallback that enforces minimum output.** The LambdaClass router falls back to Uniswap V3 ([audit update, 2026-09-22](https://blog.lambdaclass.com/lambdaclass-propamm-router-audited-version-now-live/)).

## What to skip

- **Per-level book writes.** Median 152013 gas against 26706 for a `p`/`s` write (paper).
- **Inferring a venue's strategy from its volume rank.** Most Solana leaders do not publish their model (Solana report).
- **Copying the Priority Update Registry v1 without version checks.** Flashbots disclosed a storage-aliasing/authorization flaw on 2026-09-16 that requires migration ([advisory](https://collective.flashbots.net/t/security-advisory-storage-aliasing-in-priority-update-registry-v1-09-15-26/6030)).
- **Treating markouts as user welfare.** The Blockworks markouts cover 2026-01-26/2026-02-09 Jupiter-routed SOL/USDC flow only.
- **Averaging share figures from different sources.** Every share conflict in the evidence table below is a difference in denominator or window.

## Offchain computation and gas: what cuts cost and what the limits are

**Has to be onchain (Fact, mechanism report / paper).** Curve evaluation, inventory accounting, penalty application and settlement.

**Can be offchain.** Fair value, the spread and skew model, counterparty classification, update signing, and the choice of when to update.

**Update vs swap cost (Fact, paper §7.1, to 2026-08-31)**

| Chain | Median update | Swap | Ratio |
|---|---|---|---|
| Solana | 485–676 CU | ≥16938 CU (PropAMM leg) | >10× |
| Base | 26774–32606 gas | 201243–402881 gas | ~6–15× |
| Monad | 33428–50839 gas | 162503 with access list / 330103 without (Metric) | ~5–15× |

Conflict: Helius reports about 150000 CU for a whole Jupiter transaction through a PropAMM. The paper measures only the PropAMM leg. Treat the ratio as reliable and the absolute number as scope-dependent.

**Ethereum benchmarks ([ethereum-gas sources](https://github.com/flashbots/priority-update-registry/blob/8701e8e13d74beef9126c3e1fbbe89c64ce0c0f6/snapshots/GasBenchmarkTest.json), retrieved 2026-10-26)**
- Priority Update Registry direct update (k=0): 30698 gas.
- Batched updates: n=5 costs 116008 gas (about 23202 per update, derived) and n=10 costs 207802 gas (about 20780 per update, derived).
- Cold read: 3554 gas.
- Factory mock: quote 38890, warm quote 8890, swap 65467 gas ([PropAMMGasTest.json](https://github.com/lambdaclass/propamm-factory/blob/8ded63cd09dc6ee628b17188a13ced14b24a7ea1/contracts/snapshots/PropAMMGasTest.json)).
- Single mainnet samples, which include all other calls in the transaction and are not like-for-like:
  - El Zorro `updateState`: 41986 gas.
  - Whole transactions containing a swap: El Zorro 1270043, FermiSwap 879547, Metric 334379, bopAMM 417956.
- **UNVERIFIED:** isolated El Zorro swap gas, typical LambdaClass Router end-to-end gas, and WOOFi sPMM Ethereum swap gas.

**What cuts cost.** One write per reprice. Batching updates amortizes the 21000 base cost: per-update gas falls from 30698 to about 20780 at n=10. Builders can drop updates that are not needed (Titan conditional inclusion, [docs.titanbuilder.xyz/propamms](https://docs.titanbuilder.xyz/propamms)).

**Limits (Inference).**
- Router requoting, failed simulations, builder charges and fallback execution can cancel the saving.
- The 21000 floor sets a lower bound on any standalone EVM update.
- The size of the saving for WOOFi is **UNVERIFIED** until a matched replay exists.

## Distribution: which aggregators and wallets matter and why

| Path | Status (source, 2026-10-26) | Why it matters |
|---|---|---|
| Jupiter Metis (Solana) | Live registry lists the major PropAMMs and WOOFi | Effectively the Solana entry point. Requires quote/swap parity, an audit and a traction review |
| OpenOcean | Live source lists: WOOFi and/or Metric on 9 EVM chains; Tessera, ElfomoFi and Kipseli on Base | Machine-checkable listing |
| 0x / Matcha | Dated changelog additions: Base, BSC, Robinhood Chain (Tessera, Elfomo, WOOFi V2) | 0x names Coinbase Wallet, Robinhood Wallet, Phantom and MetaMask as users of its stack ([0x docs](https://docs.0x.org/docs/core-concepts/introduction-to-0x)). Vendor claim |
| KyberSwap | Adapters exist for Tessera, Metric, ElfomoFi, WOOFi. Live chain enablement UNVERIFIED | Needs a tracker, a simulator and a calldata adapter |
| DODO X | Documented WOOFi V2 on 8 EVM chains | Listing only, no flow share |
| 1inch, Odos, OKX DEX | WOOFi listing on 1inch is historical. Odos endpoint returned HTTP 530. OKX is an RFQ path (1 s price refresh, 500 ms firm response, >90% success) | Current routing UNVERIFIED |
| LambdaClass Router (Ethereum) | Venues: FermiSwap, Kipseli, Bebop, Tempest, TaurusFi, Metric, El Zorro | The Ethereum PropAMM access path |

Flow share through any aggregator or wallet is **UNVERIFIED** in every case: all `share_of_flow_pct_if_known` cells are empty. Inference: these listings show which venues are eligible for routing, not which ones win orders. Prioritize the paths that have live, machine-checkable registries (Jupiter, OpenOcean) and those with documented wallet reach (0x).

## Ethereum specifics

- **Facts (vendor-described).**
  - Makers send signed updates that builders order ahead of trades.
  - Venues read the Priority Update Registry, and the router requotes at inclusion time with a Uniswap V3 fallback.
  - Router introduced 2026-07-14; audit update 2026-09-22; Factory announced 2026-10-14. **The starting lead that both launched in October is wrong.**
  - El Zorro is Zonda's Factory deployment.
- **The paper excludes Ethereum** (PropAMMs from May 2026) because builder ordering makes them RFQ-like (paper §2). Titan documents builder priority, freshness protection and conditional update inclusion. These are builder policy, not a protocol guarantee.
- **Volume.**
  - Tracked cohort: 2127196166 USD for 2026-09-25/2026-10-24 (5.633%).
  - Fermi + Metric: 458060000 USD for 2026-09-14/2026-09-20 (sum from the [ethresear.ch table](https://ethresear.ch/t/ethereum-propamms-today-and-tomorrow/26129)). Zonda's "over 458000000" cites the same series, so it is not independent corroboration.
  - LambdaClass's figures are not comparable with each other: 546000000+ USD cumulative (start date unknown) against about 13000000 USD per day (averaging interval unknown).
- **Risks.**
  - Dependence on builder coverage (Inference: few builders → concentration).
  - The registry v1 flaw (advisory dated 2026-09-16).
  - Stale signed updates.
  - Fallback economics are UNVERIFIED.

## Open questions ranked by decision value

1. **Total-cost replay** of the current WOOFi design against a `p`/`s`-style single-write adapter and against registry batching, on identical public trade intents. This decides whether a rebuild is worth doing.
2. **Same-pair DEX denominators** for stablecoin, tokenized-equity and long-tail pairs on Base, BSC and Robinhood Chain. This decides asset expansion.
3. **Aggregator flow share by venue** (0x, OpenOcean, KyberSwap, Jupiter). This decides which integration comes first.
4. **A PropAMM census for Sui, Sei, Aptos and Unichain.** This decides whether to take on a new runtime.
5. **Ethereum builder coverage and fallback share** for router fills. This decides whether to build an Ethereum adapter.
6. **A frozen, address-level Solana snapshot** that resolves the DefiLlama value conflicts and the Tessera/Tessera V and SolFi/SolFi V2 program mapping.
7. **Wash-adjusted and distinct-taker volume.** This affects confidence in every share figure.

## Evidence quality table (claim, source, confidence)

| Claim | Source | Confidence |
|---|---|---|
| Paper published 2026-09-29; sample 2025-09-01/2026-08-31; Base launch 2025-10-30; Monad early Apr 2026 | [arXiv 2609.38056](https://arxiv.org/abs/2609.38056) | High that the paper says this. Launch dates not independently verified |
| Blockworks "Block Building Wars" is dated 2026-10-14 | Lead. **Contradicted:** dated 2026-02-17 ([Blockworks](https://app.blockworks.com/report/solana-s-block-building-wars)) | High (correction) |
| Router and Factory both launched in Oct 2026 | Lead. **Contradicted:** router 2026-07-14, Factory 2026-10-14 | High (correction) |
| Solana PropAMM share >50% (paper) vs 26.1% (TDMM, 30d to 2026-09-26) | Paper; [TDMM](https://tdmm.io/insights/blog/top-crypto-market-makers/), a market maker, so commercially interested | Not a conflict: SOL/USDC vs all pairs. Do not average |
| Solana tracked cohort 33.954%, 2026-09-25/2026-10-24 | DefiLlama API sums (upstream) | Medium: cohort incomplete |
| Named five-venue cohort 28.083%, 32.242% with Scorch (30d to 2026-10-26) | DefiLlama chain table, rounded | Medium (estimate) |
| BisonFi 30d volume | DefiLlama: 8095000000 to 9404000000 across pages on the same day | Low: unresolved snapshot conflict |
| Solana chain DEX 30d | DefiLlama 55638000000 vs investors mirror 55563000000 (same date) | Medium: small, unexplained conflict |
| `p`/`s` write 26706 gas vs level writes 152013 | Paper decode | Medium-high (authors' measurement, not rerun) |
| Registry batch per-update gas ≈20780 at n=10 | Flashbots repo benchmark (derived) | Medium: test benchmark, not mainnet |
| WOOFi sPMM Ethereum swap gas | — | UNVERIFIED |
| Aggregator flow share per venue | — | UNVERIFIED |
| Headroom USD by chain | This memo's arithmetic from DefiLlama | Low (estimate, mechanical) |
| PropAMM presence on Sui/Sei/Aptos/Unichain | — | UNVERIFIED (absence not shown) |
| Organic vs wash volume | — | UNVERIFIED |

**Limits.** This memo synthesizes upstream extractions and does not re-pull them. It shows the inputs are internally consistent, not that the underlying data is true. An independent reviewer would be needed for that evidence.