News Feed

JW3 Academy is live: terminal, guides, and what is next

jw3.ai relaunches as a trading terminal plus a multilingual Web3 academy. What moved, what is new, and what ships next.

Tokenized stock is not stock ownership

Buying $AAPLx buys a token, not Apple. What tokenized equities actually give you, what they quietly do not, and what the regulators have and have not approved.

Your quote is not your fill

0x analyzed 84,163 Uniswap v4 hooks and found 54.2% malicious. Trades routed through them settled up to 50% below the quoted amount. What that means for the number on your screen.

You can be right and still be closed

Every major perp venue runs an ADL step after liquidation and the insurance fund. Hyperliquid ranks you on (mark/entry) x (notional/account value). dYdX picks randomly. Being profitable does not protect you.

Your liquidation price is an estimate

Liquidations trigger on mark price, not last price. Kraken says the reason plainly: last price is the cheapest number to manipulate. Hyperliquid builds mark from a three-way median, and warns its own displayed liquidation price may not be the real one.

Non-custodial is not the same as decentralised

Non-custodial is a claim about keys. It says nothing about who can halt trading, who sets an oracle, or who decides what gets listed. Uniswap lists non-custodial, non-upgradable and permissionless as three separate properties. A venue can have the first and fail the other two.

The fund is a multiple, not a balance

A futures insurance fund balance is not the loss-absorption budget. Binance defines each fund's Maximum Takeover Capacity as a confidential multiple of its balance, calibrated to a 99.9% confidence interval that the venue explicitly says nothing about how often ADL happens. Coin-margined contracts share one fund per collateral asset, which makes them smaller and more exposed.

One share almost closed a $400M market

A single SK Hynix share printed 29.96% below the prior close on a thin Seoul pre-market. The print travelled 5,500 miles into an on-chain perpetual, the mark fell 18.7%, and roughly 960 long accounts were liquidated. The risk was in the reference market, not in the crypto venue. How that happened, and what it tells you about any perp tracking an asset that does not trade 24/7.

Your signature is not a transaction

On CoW Protocol, an intent is a signed expression of trading constraints, not an executable transaction. The settlement contract enforces two rules and no more. The price comes from a bonded solver competition, and the surplus split includes protocol fees of 50% of surplus, capped at 0.98% of volume.

Protected flow still gets sandwiched

Order-flow auctions work, and the residual is measurable. A three-year six-chain measurement counts 30,607 protected-flow sandwiches on Ethereum by 7 bots, and finds that 66.8% of private sandwich victims surface in at least one auction dataset. The exposure is the auction itself.

Three architectures, three different counterparties

An AMM makes you the counterparty. A CLOB makes you a queue. An RFQ makes you an auction item. The three differ in who your counterparty is and who sets the price, and that single fact predicts fill quality better than any fee tier.

Your trade log can prove the numbers, not the reasoning

Annualising one funding payment is a guess

A dashboard's annualised funding figure multiplies one observed interval by a constant that assumes the rate never changes. It describes a rate you have seen once. Here is what the multiplier hides: the interval differs by venue, the venue sets its own carry baseline, the rate can flip sign and pay you back, and the same single sample annualises to two very different numbers depending on convention alone.

Two thirds of funding rates are a constant

About two thirds of the 234 markets on Hyperliquid report exactly the same funding rate every hour. It is not a market read: it is a hard-coded interest component, documented as 11.6% APR. What that means for every funding heatmap, crowding signal and sentiment dashboard built on the number.

234 perps, five markets

Hyperliquid quotes 234 perpetual markets. Five of them carry 79.3% of the daily volume. 56 of the 234 are delisted and settled, and they still sit in the list with a price. What the distribution of a venue's own listings says about how much of it is actually tradeable.

A peg is a mechanism, not a property

"USDC, USDT, DAI and crvUSD each claim a dollar peg. None of them holds one the way paper holds a value. Each peg is a specific mechanism, and each mechanism fails at a specific speed. This article reads four of them out of their own documents: what backs the token, who can redeem and how fast, what the oracle reads when the market is thin, and why a redemption window slower than the panic is the whole ballgame."

A recovery funded by tomorrow's fees is not a refund

Drift's post-incident updates put $295M of losses against a pool seeded with about $3.8M of remaining assets. Half the shortfall is covered by committed capital. The rest is exchange revenue, which arrives only if trading survives. What a claim on future cash flow is worth, and what it is not.

A bridge is a trust decision with a UI

Cross-chain bridging comes in two architectures. An intent model takes the bonded validator set out of your path and puts solver and relayer dependence in its place, on a slower non-atomic timeline. A canonical model gives single-transaction semantics and rests on a fixed threshold you should be able to name. The interface does not tell you which one you signed up for.

The maths of a losing streak

A 70% win rate does not protect a position sized at three times the arithmetic it can carry. Ruin is set by the size of the losing streak your size allows, not by how often you are wrong. What fixed-fraction sizing does to a drawdown, why doubling after a loss reaches zero in nine steps, what volatility targeting actually targets, and the gap risk no sizing rule touches.

Your trade took a route you never read

A route is the record of how your order actually executed: which venues filled it, in what proportions, through how many hops, and against which gas estimate. Three aggregators return it in three different shapes. Almost nobody opens it.

The leverage cap is also an order-size limit

On Hyperliquid a market's leverage cap sets a hard ceiling on the size of a single order, not just on your position. 144 of the venue's 178 live markets cap at under 10x, and that tier cannot accept a market order above $500,000. What the cap tells you before you size the trade.

56 of Hyperliquid's 234 markets are delisted, and the API still quotes them

The venue's own universe list carries 234 perpetual markets. 56 of them are delisted, settled, and closed to new orders. All 56 still return a live-looking price, and a screen built on that list ranks two of them above SOL. Here is the machine-readable tell that separates them.