Every functioning market in history has settled on a trading venue that matches its underlying supply and demand. Two questions determine which venue wins: how often do buyers and sellers show up, and does someone need to step in and provide a price. CLOBs, PropAMMs, RFQs, and onchain auctions each map to a quadrant of that answer space. The mechanism follows the market. When you see a new "primitive" being sold as a universal upgrade, the right first move is to ask which quadrant of demand it actually serves.
Why this matters for a crypto investor
DeFi cycles tend to crown one venue type as the future of trading every 18 months. AMMs were the future. Then CLOBs. Then intent-based aggregation. Now PropAMMs and onchain RFQs. The cycles are noisy enough that even sophisticated allocators end up overweighting whichever venue had the loudest narrative this quarter.
The boring truth is that none of these mechanisms are new. Every one has a TradFi analog that has been in production for decades and has settled into a niche defined by what the underlying market actually needs. Understanding that mapping gives you a stable lens for two recurring questions: which DeFi venue should win for a given asset, and which protocol tokens are positioned to capture value from real flow versus narrative flow.
Two axes that define every market
Standard market-microstructure thinking organises trading venues along two axes. Both questions sound deceptively simple. The combination determines almost everything about how an asset trades.
Axis one: is the market quote-driven or order-driven?
In an order-driven market, buyers and sellers submit their own orders. Bids and offers stack up in a public book, then match each other. The exchange is a meeting place, not a counterparty.
In a quote-driven market, a professional intermediary steps in to provide an executable price. The buyer trades against the dealer, not directly against another buyer or seller. The dealer commits its own balance sheet, internalises flows from different clients, and hedges what is left.
Axis two: is natural trading interest continuous or discrete?
In a continuous market, buyers and sellers show up throughout the day, every day. Demand is broad-based and steady. The market needs to be open and quoting at all times.
In a discrete market, trading interest arrives in bursts. Someone has a specific need, a specific size, or a specific moment. Between those bursts the market is quiet. A permanent two-sided order book is overkill, and often impossible to maintain.
The four quadrants
Combine the two axes and every trading venue in the world falls into one of four boxes:
A two-axis map of trading venues. Each quadrant has TradFi anchors that have been in production for decades and DeFi analogs that map onto them.
CLOBs: continuous, order-driven markets
CLOBs are the natural fit for standardised, popular, easily understood assets that many people want to trade or own. The asset needs enough independent buyers and sellers that the order book stays two-sided through the day without a single dealer holding it up.
Blue-chip equities fit. So do BTC and ETH spot. So do major futures like Brent crude and gold. Demand comes from individuals, ETF rebalancers, asset managers, hedge funds, corporate treasuries, and many others. Orders enter the book, interact, and produce a market price continuously.
The CLOB is the venue that crypto Twitter most often defaults to as "what good trading looks like." This is partly correct (the venue is excellent at what it does) and partly survivorship bias: the assets people argue about on social media are exactly the ones whose markets fit the CLOB profile. Most assets in the world do not.
In TI coverage: Hyperliquid's onchain CLOB is the most prominent DeFi-native example, with HYPE itself capturing value from the venue's flow. Centralized venues (Binance, Coinbase) carry the dominant share of price discovery for liquid crypto and remain the CLOB benchmark.
RFS: continuous, quote-driven markets
RFS, or Request for Stream, lets a trader execute against a constantly updating stream of executable prices from specialist dealers. The dealer does not wait for a matching customer order. It commits its own balance sheet, internalises flows from different clients, and hedges the remainder elsewhere.
The single biggest example is spot FX. The OTC FX market traded roughly $9 to 10 trillion per day in the most recent BIS triennial survey (2025). Most of that does not run through a giant order book. It runs through fast-updating streams of dealer quotes priced off internal axes and external hedging venues. In major currency pairs under normal conditions, institutional spreads can be fractions of a basis point.
Why is FX structured this way? Because the underlying demand is wholesale. Banks, asset managers, corporations, central banks, hedge funds, prop firms, and global payments flows are the real participants. Retail are FX users, not FX traders. When you buy an espresso in Italy with a sterling card, you are not posting a limit order. A dealer somewhere converts at a streamed rate.
Proactive Market Maker AMMs (PropAMMs) continuously update executable prices using external market prices, inventory, volatility, and their own risk models. They commit capital, internalise flow, and hedge externally. That is exactly the structural role of an FX dealer streaming quotes. The "AMM" in the name is misleading: PropAMMs are closer to an onchain dealer-streaming venue than to a Uniswap-style constant-product pool. PropAMM venues live primarily on Solana (BisonFi, HumidiFi, SolFi, Obric, ZeroFi) and emerging on Ethereum (FermiSwap). Hyperliquid uses a CLOB underneath rather than a PropAMM model, even where HIP-3 enables permissionless market deployment.
RFQs: discrete, quote-driven markets
RFQ markets exist when the asset is too bespoke, too illiquid, or the trade is too large for continuous quoting to make economic sense. A trader pings one or more counterparties (interface, API, or chat box), the counterparties price the trade in real time considering inventory, hedging cost, and adverse-selection risk, and the buyer chooses the best quote within a short validity window.
Examples in TradFi: corporate bonds (especially below investment grade), bespoke options, structured products, very large equity blocks. The dealer adds value by warehousing risk and finding the other side, often over hours or days. Spreads are wider than CLOB markets because the work is real.
RFQs in DeFi serve a similar function: routing a large or non-trivial trade to professional market makers who price it bilaterally rather than slicing it across thin onchain pools. 1inch Fusion, CoW Swap, Hashflow, and Native are all RFQ-flavored. None of them are competing with CLOBs for small retail trades. They compete with the slippage you would eat on Uniswap if you tried to move $5 million through it.
Auctions and brokered deals: discrete, order-driven markets
The final quadrant is where interest is not continuous and where the venue brings competing orders together at discrete moments. A rare piece of jewellery has plenty of buyers but only one seller. A company issuing a security needs to gather demand once, not run a permanent book. The market does not need to be open all day. It opens, clears, closes.
Bookbuilding for IPOs, US Treasury issuance, and traditional auction houses all live here. So do bilateral OTC deals when the asset is bespoke enough that only one realistic counterparty exists at the moment of trade (a large locked-token allocation, an illiquid bond on a quiet day).
In DeFi: bonding-curve auctions, batch auctions, fair-launch mechanisms, and bilateral OTC tooling all map here. Pendle's PT-AMM has elements of this structure because the asset (a principal token with a fixed maturity) trades against a curve that is not constantly two-sided.
The DeFi translation table
Translating the quadrant map into DeFi-native terms produces a clean reference:
| TradFi venue type | DeFi analog | Fits when |
|---|---|---|
| CLOB | Onchain CLOB (Hyperliquid, dYdX) or hybrid (Binance, Coinbase) | Many independent buyers and sellers, continuous flow, standardised asset |
| RFS (Request for Stream) | PropAMMs (FermiSwap on Ethereum; BisonFi, HumidiFi, SolFi, Obric, ZeroFi on Solana) | Continuous flow but liquidity needs a dealer balance sheet to be deep |
| RFQ (Request for Quote) | Intent-based aggregators (1inch Fusion, CoW Swap, Hashflow, Native) | Discrete large or non-trivial trades, professional market makers price bilaterally |
| Auction / Brokerage | Bonding curves, batch auctions, fair launches, OTC tooling | Discrete events, often one-sided demand, asset is bespoke or has a specific moment |
Where TI-covered assets and protocols actually sit
Real assets often trade through more than one venue at once, with primary price discovery anchored to one and secondary or specialised flow flowing through others. The honest map of TI coverage:
Liquid L1s (BTC, ETH, SOL, BNB, XRP)
Primary price discovery: centralized CLOBs (Binance, Coinbase, OKX). Secondary onchain flow: AMMs (Uniswap, Curve, Raydium, PancakeSwap) for liquidity provisioning and access. RFQ for institutional blocks. The dominant venue for an L1 spot is a CLOB because the asset is standardised, popular, and has thousands of independent participants.
HYPE
Primary price discovery: Hyperliquid's own onchain CLOB. HYPE is the unusual case in TI coverage of a token whose primary venue is itself an onchain CLOB rather than a centralized one. This is part of the structural thesis behind why HYPE captures fee flow that other L1s do not: the value of the venue accrues to the token.
Pendle (PT and YT)
Pendle's primary tokens trade through a specialised AMM that incorporates time decay. The asset (a fixed-maturity yield instrument) has continuous-but-thin flow, which sits awkwardly between the RFS and auction quadrants. The PT-AMM is closer to a discrete-with-curve mechanism than to either pure CLOB or RFS. This is an example of a market where the asset's structure forced a custom venue rather than reusing an existing one.
Curator-issued vault tokens (Morpho, Euler)
Most curated vaults do not have continuous secondary markets at all. Subscription and redemption happens against the issuer through onchain primitives. This is the "no real venue, primary issuance only" case: the asset trades discrete, against the protocol, and price formation happens at the NAV layer rather than at a venue. See vault credit risk for how redemption-tier frameworks try to solve the secondary-liquidity gap.
RWA tokens (BlackRock BUIDL, Janus Henderson JTRSY, Maple)
Primary issuance via direct subscription, secondary venues emerging through institutional-tier products like Grove Basin and Upshift Clear. Today these sit firmly in the discrete quadrant. As cross-platform aggregation matures they will likely develop RFQ-style secondary venues, not CLOBs, because the underlying asset is institutional and the natural trade size is large.
When the same asset trades through multiple venues
The framework does not claim that every asset trades through exactly one quadrant. Most assets that matter trade through several at once. The primary price discovery anchors to one venue, and the others serve adjacent needs:
- ETH: CLOB (Binance, Coinbase) for primary spot; AMMs (Uniswap, Curve) for permissionless access and DEX-aggregated routing; RFQ (1inch Fusion, CoW) for institutional blocks; auctions for protocol-specific issuance events.
- Investment-grade corporate bonds (the TradFi analog): RFQ as the primary mechanism, but the largest issues develop CLOB-style electronic trading for the most liquid segments. The asset moves quadrant as its liquidity profile changes.
- HYPE: CLOB (Hyperliquid's own onchain venue plus centralized listings on Binance and others). HIP-3 enables permissionless market deployment but still settles through the CLOB mechanism, so it does not move HYPE into a different quadrant. One primary venue type, multiple venues operating it.
The pattern is consistent: assets gravitate toward a primary venue that matches their natural demand profile, then develop secondary venues for the edge cases (large blocks, bespoke structures, illiquid moments). The mistake is treating the secondary venue as a primitive that will eventually swallow the primary. It almost never does.
Markets lead, mechanisms follow
The framework is older than DeFi by several decades. It is not novel and it is not contested in market-microstructure literature. The reason it keeps showing up in DeFi discourse is that the sector is still discovering, asset class by asset class, which quadrant each crypto-native instrument actually belongs to.
That discovery is slow. It is also the most important determinant of which DeFi protocols capture durable fee flow. A protocol building a CLOB for an asset whose demand profile is fundamentally discrete will lose to an RFQ. A protocol building a PropAMM for an asset that already has deep CLOB liquidity will struggle to displace it. The mechanism does not pick the asset. The asset picks the mechanism.
When evaluating a new DeFi venue, the diagnostic question is not "does the technology work" or "does the team ship." It is: what quadrant of demand is the venue actually serving, and does the asset's natural trading profile sit in that same quadrant? If yes, the venue has structural tailwinds even if the team executes imperfectly. If no, the venue is fighting the gravity of the asset class and will need overwhelming UX or incentive advantage to win. Most projects that fail in the medium term are not bad teams. They are right team, wrong quadrant.
Key takeaways
- Two axes determine venue choice. Continuous vs discrete demand, order-driven vs quote-driven liquidity provision. Every market in the world maps to one of the four resulting quadrants.
- The mechanism follows the market. Assets gravitate toward the venue that fits their natural supply-demand structure. Technology can improve the venue. It cannot change which quadrant the asset belongs to.
- PropAMMs are the RFS analog. Not a new primitive. The TradFi equivalent has run the global FX market (roughly $9 to 10 trillion per day) for decades.
- Onchain RFQ is the institutional-block analog. 1inch Fusion, CoW Swap, Hashflow, and Native are competing for trades that should never have hit an AMM in the first place, not for routine retail flow.
- The right diligence question. Before evaluating whether a DeFi venue has good UX, good incentives, or good token economics: ask which quadrant of demand it serves and whether that quadrant exists at scale for the asset class.
This framework does not change weekly. Asset-by-asset coverage in TI research applies it to specific tokens (HYPE, Pendle, Morpho, Ethena, others). Use this concept page as the anchor when reading any of those.