The House Edge

Appearance

Chapter 2 — Research Note (Phase 1) ​

Status: RESEARCH (Phase 1 — written by agent, awaiting Nick's review) Date: 2026-05-13 Working chapter: 02 — Where Liquidity Lives Carries forward from: SPEC.md (strawman drafted by agent; Nick edited or accepted before this Phase 1 ran)

How to read this note ​

Every claim listed below is sourced. Two things to know up front:

  1. The seminal "passive LPs lose money" studies have not been re-measured for 2026. The CrocSwap (2022) and Topaze/Bancor (2021) findings still anchor the empirical picture. The chapter will frame them explicitly as the most recent comprehensive measurements. The Loss-Versus-Rebalancing (LVR) framework from Milionis–Moallemi–Roughgarden–Zhang is the live academic literature and gives the chapter its strongest theoretical claim.
  2. Solana's spot-liquidity story has changed since the chapter was first scoped. The SPEC framed it as "Solana converged on AMMs + aggregators." The 2026 data tightens that to "prop-AMMs routed through Jupiter." Anonymous prop shops (HumidiFi, SolFi, Tessera) now account for more than half of Solana DEX spot volume. The CLOB experiment is effectively over — Phoenix has been rebranded "Phoenix Legacy" by its own team. This is the chapter's biggest structural update and is flagged in Q1 below.

1. Key claims ​

Each numbered claim is something the chapter is allowed to state. Sources are cited inline.

Liquidity, AMMs, CLOBs

  1. Liquidity is the supply of resting orders or pooled inventory ready to fill a trade at a known price. Without it, a market has only intent. The two architectures that supply on-chain liquidity are automated market makers (AMMs), in which a pool of paired assets is priced algorithmically, and central limit order books (CLOBs), in which discrete bids and asks are matched.

  2. Uniswap V3's concentrated-liquidity design (the dominant AMM model on Ethereum since 2021) was designed to make passive LPs more capital-efficient than the constant-product V2 design. The empirical record is mixed: across the period studied by Uniswap Labs' own research, passive V3 positions outperformed V2 on fee returns by ~54% on average — but in the highest-volume ETH/USDC 5-bps tier, passive V3 LPs underperformed V2 by 68%. (Uniswap Labs Research — When Uniswap v3 Returns More Fees for Passive LPs, 2023) The most-traded pool is the worst case for the passive LP.

The LP economics — what the chapter argues

  1. The most recent comprehensive measurement of passive-LP outcomes on Uniswap V3 — across 17 pools representing ~43% of total V3 TVL — found that LPs collectively would have been better off holding their underlying assets, by approximately $60.8 million across the study window. About 49.5% of measured LPs realized negative returns net of impermanent loss. (Topaze Blue / Bancor — Impermanent Loss in Uniswap V3, 2021) The chapter frames this as the most recent comprehensive measurement; no 2025–2026 update has been published.

  2. The Loss-Versus-Rebalancing (LVR) framework reframes the LP problem more sharply than impermanent loss does. An LP isn't competing against HODL — they're competing against an arbitrageur who would close the same pool position via CEX rebalancing. Under LVR, a Uniswap V2-style ETH/USDC pool at 5% daily volatility loses ~3.125 bps/day (~11% annualized) to the arbitrageur, before fees. LVR scales with the square of volatility. For fees to compensate, they must cover LVR — not IL. (Milionis, Moallemi, Roughgarden, Zhang — Automated Market Making and Loss-Versus-Rebalancing, 2022 onward)

  3. Concentrated liquidity is widely under-utilized. In the canonical study of the V3 ETH/USDC 0.3% pool, 40–50% of positions were out-of-range at any moment, and only 50–60% of positions stayed in-range across an 8-month window. Roughly 15–25% of total LP capital sat idle. (CrocSwap — Impermanent Loss and JIT Liquidity in the Uniswap ETH/USDC 0.3% Pool, 2022) No 2025–2026 re-measurement of V3/V4 in-range share has been published; flagged as a research gap.

The market maker — and where MMs actually live

  1. Professional market making on most on-chain venues is concentrated in a small number of firms. On UniswapX (Ethereum), two solvers — SCP and Wintermute — together account for over 90% of volume. On 1inch Fusion in Q2 2025, the top three solvers (Rizzolver, Flowmatic, 1inch Labs) controlled roughly 68% of volume. (Khakhar et al. — Execution Welfare Across Solver-based DEXes, arXiv:2503.00738, 2025-03; Messari — State of 1inch Q2 2025)

  2. On Hyperliquid specifically, the named professional MMs are now part of the regulatory record. Wintermute operates across 76 Hyperliquid markets with approximately $199 million in resting notional and ~1,700 active orders as of January 2026. Bitwise's April 2026 amendment to its Hyperliquid ETF S-1 names Wintermute and Flowdesk as approved trading counterparties. (CoinShares — Hyperliquid: the Xetra of digital finance?, Q1 2026; Crowdfund Insider — Bitwise Submits Latest Amendment to Hyperliquid ETF, 2026-04)

Solana — the AMM-plus-aggregator stack, more sharply

  1. The Solana spot-DEX architecture in 2026 is not "AMMs + aggregators." It is prop-AMMs routed through aggregators. HumidiFi, SolFi, and Tessera — proprietary, non-public, market-maker-operated AMMs — together account for more than 50% of all Solana spot DEX volume. HumidiFi alone holds roughly 65% of the prop-AMM segment and derives over 95% of its volume from aggregator routing. (Blockworks Research — Solana DEX Winners: All About Order Flow, 2026-01-05; Helius — Solana's Proprietary AMM Revolution, 2025)

  2. Jupiter's share of the Solana aggregator market was 93.6% in early Q1 2026, with brief dips below 50% in late 2025 when Titan and DFlow gained share. Over 70% of all Solana DEX volume flows through aggregators — the aggregator has become the trading interface, displacing direct-to-pool routing for retail. (SolanaFloor — Jupiter Reclaims Dominance with 93.6% Market Share, Q1 2026; SolanaFloor — Aggregator share crosses 70%, late 2025)

  3. The most credible on-chain CLOB attempt on Solana — Phoenix, built by Ellipsis Labs — has been wound down. Cumulative lifetime spot volume reached $75 billion against cumulative lifetime revenue of $14.54 million. Q2 2024 quarterly revenue peaked at $3.70 million. Q1 2026 quarterly revenue was $68,604 — a roughly 54× decline from the peak. Ellipsis has rebranded the venue "Phoenix Legacy" and pivoted to perpetuals; the same team simultaneously runs SolFi, a proprietary AMM. (DefiLlama — Phoenix, retrieved 2026-05-13; Ellipsis Labs — Introducing Phoenix Perpetuals, 2025–2026)

  4. Solana's total spot DEX volume in Q1 2026 was approximately $284.5 billion (down 17% QoQ). The largest public AMM operators by Q1 2026 protocol revenue were Meteora ($11.4M), Raydium ($5.6M), and Orca ($1.8M). Pump.fun captured roughly 90% of Solana launchpad share with $119.8M in launchpad revenue. (Pine Analytics — Meteora Q1 2026 Quarterly Report)

Hyperliquid — the CLOB that worked, and why

  1. Hyperliquid handled approximately $177.66 billion of perpetuals volume in the 30 days ending April 2026, representing roughly 70% of all on-chain perpetuals share. Spot volume over the same window was $3.109 billion — less than 2% of the perp number. Q1 2026 perpetual fees were $137.36M against spot fees of $2.31M. Annualized fee revenue exceeded $700 million. (Yellow — Hyperliquid Owns 13% of All Perp Volume, April 2026; DefiLlama — Hyperliquid)

  2. Hyperliquid's protocol-owned LP vault — HLP — has TVL of approximately $391 million as of May 2026. Independent analysis through mid-2025 reported a lifetime CAGR of about 42% (trailing-12-month ~22%), with a Sharpe ratio above 2.89 and a maximum drawdown of approximately 6.6%. (DefiLlama — Hyperliquid HLP; Geronimo / Medium — A Risk & Return Analysis of Hyperliquid's HLP Vault, 2025-02; OnChainTimes — Analyzing HLP & JLP Returns, mid-2025) Flag: most recent independent return analysis is early-to-mid 2025; 2026 APR not published independently.

  3. The canonical adversarial-selection case for HLP is the JELLY exploit of March 2025. An attacker pumped JELLY's spot price ~400% in an hour, forcing HLP to inherit a $4.1M short. Unrealized HLP losses reached $12 million before Hyperliquid validators delisted the asset. The attacker net withdrew approximately $6.26 million. The incident triggered tighter leverage rules and remains the cleanest single case study of what happens when a protocol-owned LP absorbs targeted adversarial flow. (OAK Research — Hyperliquid JELLY Attack, 2025-03)

Ethereum and its L2s — the both-and case

  1. Solver-mediated trading has displaced direct AMM trading for a meaningful share of Ethereum retail flow. CoW Swap broke $9 billion in monthly volume in July 2025 — a single-month all-time high. CoW and 1inch each ran around $2 billion per week in Ethereum solver volume in May 2025. Within UniswapX, two firms (SCP and Wintermute) supply >90% of fills. Within CoW, the largest solver (Barter) supplied 28.2% of fills in September 2025 and was targeting >50% market share. (Blockworks — Barter buys rival solver codebase, 2025-09; arXiv:2503.00738 above)

2. Numbers to verify ​

Every figure the chapter might use, with status flags. Bold = chapter is likely to lean on it. Italic = optional color.

#NumberSourceDateFlag
N1Across 17 V3 pools (~43% of TVL), passive LPs collectively underperformed HODL by ~$60.8M; ~49.5% of LPs net negative after ILTopaze/Bancor via CryptoSlate2021Foundational; no fresher comprehensive measurement exists. Frame explicitly.
N2Passive V3 LPs in the ETH/USDC 5-bps tier underperformed V2 fees by 68%; +54% on average across all tiersUniswap Labs Research2023Strong primary; the 5-bps finding is the chapter's most pedagogical number
N3Under LVR, an ETH/USDC pool at 5% daily volatility loses ~3.125 bps/day (~11% annualized) to the arbitrageur, before feesMilionis–Moallemi–Roughgarden–Zhang2022 onwardAcademic gold standard for LP loss; LVR scales with vol²
N440–50% of V3 positions out-of-range at any time; ~15–25% of total LP capital idleCrocSwap2022Frame as foundational; not re-measured for V4
N5HumidiFi + SolFi + Tessera together >50% of Solana spot DEX volume; HumidiFi alone ~65% of prop-AMM share, >95% aggregator-sourcedBlockworks Research; HeliusQ1 2026 / mid-2025The 2026 Solana liquidity story
N6Jupiter ~93.6% of Solana aggregator share in early Q1 2026 (with dips below 50% in late 2025)SolanaFloorQ1 2026Solid; aggregator share is volatile, cite a range
N7>70% of all Solana DEX volume routed through aggregators in late 2025 / early 2026SolanaFloorlate 2025
N8Phoenix Spot Q2 2024 quarterly revenue $3.70M → Q1 2026 revenue $68,604 (≈54× decline); cumulative lifetime spot volume $75B against $14.54M cumulative revenueDefiLlama / Phoenix2024–2026Strong; the Phoenix decline curve is the chapter's anchor for "the CLOB experiment ended"
N9Manifest 30-day Solana CLOB spot volume $3.915B (May 2026); 7-day $1.157BDefiLlama / Manifestretrieved 2026-05-13Only Solana CLOB with material 2026 spot activity; still <2% of Solana spot
N10Solana total spot DEX volume Q1 2026 = ~$284.5B (down 17% QoQ); Q1 2026 protocol revenue: Meteora $11.4M / Raydium $5.6M / Orca $1.8MPine Analytics — Meteora Q1 2026Q1 2026Strong primary
N11Pump.fun Q1 2026 launchpad revenue $119.8M, >90% of Solana launchpad sharePine Analytics — Meteora Q1 2026Q1 2026Color; could be a one-sentence cameo
N12Hyperliquid 30-day perp volume ~$177.66B (April 2026); ~70% of on-chain perp share globally; spot $3.1B (1.75% of perps); Q1 2026 perp fees $137.36MYellow; DefiLlamaApril 2026Strong
N13HLP TVL $391M (May 2026); lifetime CAGR ~42%; TTM CAGR ~22%; Sharpe ~2.89 lifetime; max drawdown ~6.6%DefiLlama HLP; Geronimo2025–2026TVL is current; return profile is most recent independent (2025)
N14JELLY exploit (March 2025): attacker pumped spot 400%; HLP inherited $4.1M short; peak HLP unrealized loss $12M; attacker net $6.26M; validators delistedOAK Research2025-03Canonical HLP adverse-selection case
N15Wintermute on Hyperliquid: 76 markets, ~$199M resting notional, ~1,700 active orders (Jan 2026); Wintermute + Flowdesk named in Bitwise April 2026 ETF S-1CoinShares; Crowdfund Insider2026-01 / 2026-04Strong named-actor anchor for the MM sidebar
N16Hyperliquid taker fee 4.5 bps base on perps (was 3.2 bps in earlier research; corrected per Ch 1 Phase 1 research)Hyperliquid Fees docsliveThe only published HL fee/spread datum I could pin
N17CoW Swap July 2025 single-month volume >$9B (ATH); CoW + 1inch ≈$2B/week each, May 2025; SCP + Wintermute >90% of UniswapX volume; Barter at 28.2% of CoW solver share in Sept 2025, targeting >50%Blockworks; arXiv:2503.007382025The solver-share story; concentration is the through-line
N18Uniswap V3 TVL ~$2.118B (May 2026); 30-day volume ~$15.42B; annualized fees ~$206.4M; Q1 2026 protocol revenue $95.39M; V4 TVL $805MDefiLlama Uniswap V3retrieved 2026-05-13Current state of Ethereum's dominant AMM; useful in chain comparison

Deliberately not included:

  • A current "% of Ethereum retail flow that is solver-mediated" headline number — no single citation found for Q1 2026; directional evidence points to >30% but the chapter will not assert the exact figure.
  • A Hyperliquid BTC perp top-of-book spread in basis points — not published in any source I could locate. The chapter will either omit it or cite the taker fee (3.2 bps) as the only public number.
  • A 2026 re-measurement of Uniswap V3/V4 in-range share — does not exist.

3. Contested or evolving claims ​

  • Whether passive LPs still lose money in 2026 — and whether they ever did. The Topaze/Bancor (2021) and CrocSwap (2022) findings have been the dominant empirical claim. Critics argue (a) the studies pre-date most active-LP management tooling, (b) the analysis was loss-versus-HODL rather than loss-versus-rebalancing, which underweights LVR. The LVR framework gives a cleaner and probably worse answer for the LP. The chapter should frame the 2021–2022 numbers as the most recent comprehensive measurement and the LVR framework as the live theoretical lens.

  • Whether the "prop AMM displaces public AMM" story generalises. The Solana evidence is strong: HumidiFi, SolFi, Tessera have demonstrably captured the spot venue. On Ethereum, the equivalent story is solver-mediated trading (UniswapX, CoW). The mechanism is different — Ethereum's solver markets compete to fill intents, not to host liquidity — but the consequence is the same: the passive-LP pool is increasingly the back-stop that informed flow trades against, not the front of the market. The chapter can argue this; it should be careful to distinguish the two mechanisms.

  • Whether Hyperliquid's CLOB success is portable. Hyperliquid handles perpetuals; its spot is ~1.75% of perps. Every prior on-chain CLOB attempted spot. The honest framing is that Hyperliquid built a CLOB that worked for perpetuals, not "the CLOB that worked." Spot CLOB on-chain remains, in 2026, a solved problem only on Hyperliquid (and only for a small fraction of its volume) and on Manifest (a small fraction of Solana spot). The chapter should not over-generalise.

  • The HLP return profile in 2026. Lifetime CAGR of 42% (from early-2025 analysis) is impressive; 2026 fee accruals are sharply lower (Q2 2026 partial = 62% decline QoQ). The 2026 annualised return is likely materially lower than the lifetime number suggests. The chapter should cite the lifetime/TTM number with the explicit caveat that 2026 fee revenue has contracted.

4. Characters introduced ​

The Market Maker (full sidebar — first appearance of the book's "Meet the actor" format)

  • Who they are. A firm or operator that supplies two-sided liquidity to a venue, earning the spread between bids and asks. On-chain, MMs come in two forms: off-chain MMs quoting on-chain (Wintermute, Flowdesk, Selini quoting on Hyperliquid; the same firms operating Ethereum solver wallets) and on-chain prop-AMM operators (Temporal-operated HumidiFi, the Ellipsis-operated SolFi, others).
  • How they earn. The spread, the rebate from venue makers' programs, and — for solver-based venues — the difference between the price they fill the user at and the price they hedge at on the next venue (typically a CEX).
  • How they spend. On inventory risk (the position they're stuck with after each fill until they can hedge), on infrastructure (low-latency routing, co-located hardware), and on the moat that increasingly defines the business: relationships with the venues whose flow they fill.
  • TradFi analogue. A market-making desk at Citadel Securities, Jane Street, Optiver, or DRW — the same role on equity exchanges, slightly different mechanics.

The Passive LP (no sidebar, but a half-page "How an LP earns and loses" treatment)

  • A retail or institutional user who deposits assets into an AMM pool and earns a share of trading fees in exchange for absorbing impermanent loss (and, against the more recent LVR framework, a continuous loss to the arbitrageurs who keep the pool's price aligned with the broader market). The reader meets the passive LP for the first time here; the role is referenced retrospectively in Chapter 4 §5a and §7 as "the silent counterparty to CEX-DEX arbitrage."

The Aggregator — Jupiter as the named exemplar

  • A protocol that takes a trader's intent ("swap $10K of USDC for SOL") and routes it across multiple venues to minimise the trader's effective slippage. Jupiter dominates Solana (~93.6% Q1 2026); on Ethereum, the equivalent layer is fragmented across 1inch, CoW, 0x, Paraswap, and direct Uniswap. The chapter treats Jupiter in cameo here; the full infrastructure treatment is Chapter 6.

The Prop AMM Operator — Temporal (reportedly operating HumidiFi)

  • A new category, introduced briefly: a small market-making firm running a private AMM-style execution venue whose liquidity is non-public but whose settlement is on-chain. Reportedly Temporal for HumidiFi; Ellipsis Labs runs SolFi; Tessera is the third major operator. The chapter develops the prop-AMM idea enough to ground the Solana convergence story.

HLP — cameo for the Hyperliquid section

  • Hyperliquid's protocol-owned liquidity vault. Mechanically, an AMM-of-MMs that absorbs liquidation surplus and quotes against incoming flow. Functions as the protocol's house book. Full mechanical treatment in Chapter 9; appears here in the chain-comparison box.

The Solver — cameo only

  • The Ethereum-side equivalent of a prop AMM: a firm that bids to fill a user's transaction intent on-chain, sourcing liquidity from any venue it chooses. Full treatment in Chapter 7 (Exclusive Order Flow).

5. Worked example candidates ​

The SPEC offered three options. The 2026 research reshapes the comparison.

Candidate A — Bob's $50,000 LP on Uniswap V3 ETH/USDC ​

A retail business reader — call him Bob, a friend of Alice from Chapter 4 — deposits $50,000 into the Uniswap V3 ETH/USDC 5-bps pool in February 2026. The chapter tracks his position over a month: fees earned, IL accrued, the LVR-equivalent benchmark. The 5-bps tier finding from Uniswap Labs (passive LPs underperform V2 by 68% in this tier) plus the Topaze/Bancor LP-loss finding (49.5% of LPs net negative) anchor the math. Endpoint: Bob lost more to the arbitrageurs than he earned in fees. Not by a lot — but enough that the chapter can show the dollar.

  • Pros: relatable, dollar-anchored, lets the chapter develop the LP role in detail, plants the "passive LP as chronic loser" thread that Chapter 4 already referenced. Strong forward link to Chapter 11.
  • Cons: requires the chapter to develop LVR plus IL plus concentrated liquidity in close succession — there is a real risk of the reader drowning in mechanics. The Bible's "diagrams replace prose" rule applies hard here.

Candidate B — A market maker on Hyperliquid posts a BTC perpetual quote and gets adversely selected ​

Wintermute (named, real) is quoting a two-sided BTC perp market on Hyperliquid. A CEX prints first — say, Binance moves the BTC mark by 5 bps. A taker on Hyperliquid lifts Wintermute's stale offer. Wintermute now holds an inventory position they didn't want and has to hedge. The chapter walks the dollar: the spread Wintermute collected, the inventory hit, the hedge cost. HLP's role as the back-stop appears here — when Wintermute can't or won't quote, HLP fills.

  • Pros: introduces Wintermute as the named actor for the MM sidebar; develops adverse selection concretely; sets up the chain-comparison observation that Hyperliquid's MMs survive because HLP absorbs the worst flow.
  • Cons: requires the chapter to develop perpetuals briefly; risk of becoming a Hyperliquid-centric chapter when the SPEC wanted balance.

Candidate C — Alice's $10K swap from Chapter 4, replayed through the aggregator routing ​

Alice's $10K USDC→SOL swap from the Chapter 4 prologue. The chapter shows the Sankey-style routing diagram: 60% through HumidiFi via Jupiter, 25% through Orca, 15% through Meteora. Each leg gets a price. The chapter uses this to introduce the aggregator and the prop-AMM displacement of public LPs in one frame.

  • Pros: continuity with Chapter 4; shows the architecture in motion; lets the chapter introduce HumidiFi and Temporal in passing.
  • Cons: thin on LP economics — Bob's LP P&L is the chapter's strongest pedagogical anchor.

Agent's recommendation ​

Candidate A (Bob's LP P&L) as the anchor, with B as a half-page cameo in the chain-comparison box and C as the opening setup for the aggregator section. Rationale: A is the chapter's most adversarially-relevant story (the passive LP is the role Chapter 11 returns to), it carries the Bible's "show the dollar" mandate cleanly, and the 5-bps tier finding gives the reader a vivid number to walk away with. B and C are too valuable to cut entirely but neither is strong enough to anchor.

6. Open questions for Nick ​

Q1 — The Solana convergence framing is sharper than the SPEC's. The SPEC said "Solana converged on AMMs + aggregators." The 2026 research narrows that to "Solana converged on prop-AMMs routed through Jupiter — the public AMM business is real but the dominant liquidity is private." This is a more interesting and more adversarial claim than the SPEC anticipated. Do you want the chapter to develop the prop-AMM thread fully, or save it for Chapter 8 (Solana) and keep Ch 2's framing closer to the SPEC's?

Q2 — The LP-loss data is 2021–2022 vintage. The most recent comprehensive empirical measurement of passive-LP P&L is from 2021 (Topaze/Bancor) and 2022 (CrocSwap). No 2025–2026 update has been published. The chapter will frame the older numbers as "the most recent comprehensive measurement" (same treatment Chapter 4 used for the $233.8M CEX-DEX figure). The live academic work is now the LVR framework, which gives a sharper theoretical answer but no empirical 2026 dataset. Acceptable, or should I ask the research to look harder for a 2026 LP-loss measurement?

Q3 — Naming Temporal as HumidiFi's reported operator. DL News reporting attributes HumidiFi to Temporal but the operator has not publicly confirmed. The chapter could (a) name Temporal with the "reportedly" hedge, (b) say "an anonymous prop shop," or (c) skip the attribution entirely. My recommendation: (a). The book's voice tolerates "reportedly" for an on-the-record attribution that the named party hasn't denied.

Q4 — The Phoenix decline as the canonical "CLOB experiment ended" anecdote. The 330× quarterly-revenue decline plus the team's own rebrand to "Phoenix Legacy" is a clean story. Do you want this as the chapter's opening cold open, the closing point of the Solana section, or a footnote? My instinct: cold open or near-cold-open — it's the chapter's most vivid single image.

Q5 — How much Hyperliquid here vs. Chapter 9. Wintermute on 76 markets, HLP's $391M TVL, the JELLY exploit — all load-bearing for Chapter 2's chain-comparison box and the MM sidebar. But Chapter 9 is the dedicated Hyperliquid chapter. Worth confirming the budget here: my plan is to use Hyperliquid in cameo (the architecture works, here's why) and reserve the JELLY exploit and the full HLP mechanics for Chapter 9. Acceptable?

Q6 — Worked example confirmation. Candidate A (Bob's $50K LP P&L) as the anchor; B (Wintermute adverse selection on Hyperliquid) as a chain-comparison cameo; C (Alice's swap routed) as the aggregator-section setup. Confirm or override.

Q7 — Hyperliquid spread / lead-lag data isn't published. I could not find a Q1 2026 BTC perp top-of-book spread number or a peer-reviewed price-discovery study comparing HL vs Binance prints. Options: (a) omit, (b) cite only the 3.2 bps taker fee, (c) commission a Dune query before Phase 3 (slow). My recommendation: (b) plus a footnote noting the data isn't published.

Sources cited ​

Primary research:

Aggregators / news (cited because primary not available):

Phase 1 is complete. No prose, outline, or draft will be produced until Nick reviews this note and approves it (with any changes) for Phase 2.