Half I requested whether or not an AMM floor fashioned. Half II opened the pool. This piece seems on the different state object: posted depth in a CLOB.
Hyperliquid turned exhausting to disregard for multiple purpose.
HYPE’s value motion helped draw consideration. I don’t deal with token efficiency as market-structure proof. Charges are the cleaner hook: in a 2026 H1 DeFiLlama pull (Jan 1–Jun 30 UTC), Hyperliquid logged about $419 million, forward of Aave and Uniswap in the identical peer set.
Reuters, citing CryptoQuant, additionally named Hyperliquid amongst main offshore perp venues in a report on 2025 perp quantity: $61.7 trillion globally, towards $18.6 trillion in spot. I’m not merging that headline determine with DeFiLlama each day charges; the scopes differ. It solely explains why this isn’t a random venue decide.
The context is noisy. If marginal capital is crowding into AI, a number of high-attention names, or leveraged venues, weak spot exercise can appear like an AMM drawback earlier than it’s one. Charges make Hyperliquid value inspecting. They don’t determine the pool-versus-book query.
When the market types with no pool, what state object replaces it?
Hyperliquid is the case examine as a result of its public data API exposes sufficient guide state to make that query answerable.
This isn’t an AMM demise declare
“AMM is dying” bundles too many claims-volume share, LP economics, launch patterns, dealer consideration. I’m not adjudicating all of them.
The helpful query is narrower: what market object matches the product?
Callable spot nonetheless maps to swimming pools. Perps want margin, funding, liquidation, open curiosity, and place accounting-state a passive curve doesn’t carry by itself. That break up is previous in TradFi. What’s newer on-chain is inspectability: payment stream giant sufficient to note, plus public APIs that return guide, funding, and OI fields.
For an AMM, the audit begins with reserves and swaps. For a CLOB, it begins with bids, asks, fills, funding, open curiosity, and no matter slices of the chance engine leak out. Price rank doesn’t settle that break up.
When a non-AMM venue each collects actual charges and publishes l2Book, the guide stops being solely a UI widget. It turns into a partial market object-imperfect, capped at 20 ranges per facet in Hyperliquid’s case, however callable from outdoors.
Callable stock vs posted depth
An AMM and a CLOB fail otherwise.
An AMM exposes callable stock ruled by a pricing curve. A pool holds reserves and quotes by means of a rule reminiscent of x * y = ok. A dealer doesn’t select a counterparty. The pool is the counterparty. That’s the reason Half II may measure execution by operating pool state forward-change commerce dimension and execution drag modifications; change reserves and executable depth modifications; transfer the exterior value and arbitrage pushes the pool till the remaining hole is fee-limited.
A CLOB exposes persistent guide depth by means of price-level commitments. There is no such thing as a single reserve pair that defines the market. There are bids, asks, sizes, spreads, and depth across the mid. A taker consumes posted orders. A maker can cancel, repost, or transfer stock elsewhere.
AMM liquidity is dedicated right into a contract. CLOB liquidity is posted right into a queue.
The benefit isn’t {that a} guide clears what a pool can not. The guide’s native match is conditional liquidity: at this value, for this dimension, for now. A maker can tighten, widen, cancel, or hedge as stock and danger change. Queue place and timing turn out to be a part of the market.
A pool could make its protection layer extra lively by means of ranges, charges, and hooks, but it surely nonetheless exposes stock by means of a rule. It doesn’t turn out to be a local price-time queue.
Within the AMM panel, the seen state is compact: a reserve level plus a pricing rule. Within the CLOB panel, the seen state is a ladder of value ranges. Every stage can comprise a number of posted orders ( n within the API response is order rely at that stage, not queue place). The guide can change with no commerce, as a result of makers can cancel or substitute orders.
A small Hyperliquid snapshot
On 2026–06–30T02:09:33.700Z, I pulled l2Book snapshots for 4 perps on the first Hyperliquid dex and computed mid, unfold, and visual notional depth inside ±1% of mid. At snapshot time, spreads had been tight- 0.17 bps on BTC, 0.63 on ETH, 0.13 on SOL, 0.15 on HYPE-and seen depth inside ±1% of mid ranged from sub-million on SOL to roughly $18-23M on ETH (see Fig. 3).
The general public l2Book endpoint returns at most 20 value ranges per facet. For all 4 markets on this snapshot, each seen stage fell contained in the ±1% band-so ±5% depth equals ±1% depth within the processed file. That could be a measurement cap, not proof that depth is flat past 1%.
The snapshot is a point-in-time view of what the general public guide uncovered when the probe ran-not a declare about everlasting liquidity.
Depth right here comes from posted ranges across the mid, not from reserves on a curve. In Half II, depth got here from the pool state. Right here it comes from the guide.
The identical probe window additionally pulled funding and open curiosity from metaAndAssetCtxs—fields with no direct AMM analog. BTC at snapshot time: funding 0.0000125, open curiosity 35,637 contracts (base models), 24h notional quantity $3.12B; mark 60,652, mid 60,652.50, oracle 60,654 (premium −0.000016).
What the guide exposes — and what it does not
Every mechanism leaves the outsider a distinct audit drawback.
A public CLOB feed can expose extra quick construction than an AMM pool:
- greatest bid and greatest ask (l2Book)
- unfold and visual depth round mid (derived from l2Book)
- current fills (recentTrades – brief window; 10 trades returned for BTC on this probe)
- funding and open curiosity (metaAndAssetCtxs)
- 24h notional quantity (dayNtlVlm)
- payment schedule (documented in API; not re-derived right here)
The lacking facet remains to be necessary.
The guide reveals posted depth, not why it was posted, whether or not a maker intends to remain, or queue precedence. On a perp venue, margin and risk-engine state sit alongside the ladder. Liquidation resolution paths are solely partly seen; the remainder stays outdoors the artifact.
CLOB knowledge is granular and incomplete on the identical time.
Dynamic charges change the AMM protection layer
AMMs are usually not standing nonetheless. Uniswap v4-style hooks and dynamic-fee swimming pools make that clear. The previous pool design handled the payment as a largely mounted parameter. The newer design area lets the pool react: a hook can replace charges earlier than a swap, add customized accounting, or alter conduct across the swap path. Official docs place v4 hooks as a strategy to customise pool behavior-not as a path to show the pool into an order guide.
That issues as a result of many AMM failures land on the LP facet. The pool provides stock by means of a deterministic rule. If the surface value strikes first, knowledgeable stream can commerce towards stale stock. If liquidity is concentrated within the unsuitable vary, the LP absorbs the repricing. If routing or MEV makes the stream poisonous, the payment could not compensate the stock danger. Half II measured that pressure by means of execution drag, arbitrage residual, and fee-vs.-LVR.
Dynamic charges and hooks are one response-Half IV will run that protection layer in simulation. Right here the purpose is narrower.
Dynamic charges are usually not an AMM victory lap. They’re proof that the previous fixed-fee pool was under-defended towards poisonous stream.
They don’t make the pool right into a CLOB. They make the pool’s protection layer extra lively.
On a CLOB, liquidity is already active-makers quote, cancel, repost, hedge, or depart. Failure reveals up in stale quotes, queue video games, and liquidation parameters greater than in a passive curve being picked off.
The place loss lands will depend on the mechanism. In an AMM, protection more and more sits in charges, hooks, and LP technique. In a CLOB, it sits in maker conduct, margin guidelines, and liquidation logic-and a lot of that’s solely partly seen from outdoors.
The chance floor modifications too
Half II stayed inside AMM mechanics: whether or not charges cowl passive stock being repriced by arbitrage.
AMM failure tends to land on LP stock -stale costs, LVR, vary placement, poisonous stream, sandwich and routing results, and whether or not hooks or dynamic charges truly defend the pool.
CLOB failure tends to land on lively market making -queue precedence, cancellation latency, unhealthy quotes, liquidation cascades, and risk-engine parameters. A perp CLOB provides funding, open curiosity, margin, and liquidation on high of the guide.
Fastened-fee swimming pools are high quality for long-tail spot till volatility or poisonous stream arrives. Dynamic charges assist provided that the oracle or proxy is true and the hook isn’t gamed. Books work when makers keep; they break when makers depart or the chance engine misfires.
AMM and CLOB expose totally different slices of the market and lose cash in other places. Each depart one thing outdoors the artifact.
Closing
A market can kind with out an AMM pool if there may be posted depth, matching, and danger infrastructure behind the guide. I might not learn this probe as industry-wide migration. It reveals what turns into seen when the forming object is book-shaped.
The book-shaped facet is inspectable sufficient to audit-incomplete, however not hypothetical. That modifications the place loss lands and which failures keep contained in the artifact.
The query I hold returning to: the place does the mechanism fail, who absorbs the loss, and what can an outsider confirm?
For an AMM, begin with reserves and swaps, then charges and hooks. For a CLOB, begin with the guide and no matter funding/OI the API returns; assume maker intent and liquidation paths are partly lacking.
The pool isn’t all the time the article. The audit drawback stays.
Appendix: supply
- Half I: The token appeared twice. The AMM market fashioned as soon as.
- Half II: Earlier than MEV, I Constructed a Rust AMM Lab to Measure Pool State.
- Reproducibility gist — DeFiLlama charges and Hyperliquid data probe.
This put up was initially revealed on my private weblog: https://egpivo.github.io/2026/07/19/when-the-market-forms-without-an-amm-pool.html.
Hyperliquid Exhibits What a Market Seems Like With out an AMM Pool was initially revealed in The Capital on Medium, the place persons are persevering with the dialog by highlighting and responding to this story.