妙蛙种子 | CapitalismLab
妙蛙种子 | CapitalismLab|Oct 16, 2026 11:38
PAPER's loss making mining has only been popular for less than a week, but Jupiter quickly launched 1000x leverage to ride on the heat Everyone is calling it a casino, but I think this may be the most honest product in this round of Perp DEX because it's even called YOLO Let's first look at the mechanism: 1. The target is only SOL, BTC, ETH, with a maximum of 1000x and a starting price of $10. It can only be a market order without stop loss 2. It is not a peer-to-peer match, the price is based on the real market price, but all wins and losses are settled in a House pool, and the House's money comes from the trader's losses 3. A 0.1% fluctuation under 1000x will directly reset to zero When the House pool is not enough, the winner's profit has to queue up and there is no guaranteed redemption time 5. Beta stage, unaudited Essentially, Jupiter is not an exchange, but a market maker. What the user bought is not BTC directional exposure, but a lottery ticket that becomes invalid with a 0.1% fluctuation. If they win, it depends on whether the banker's pool is enough to compensate Looking at PAPER again: Papertrade-xyz on Hyperliquid has been operating at a loss for less than 6 days, and the three addresses with the highest transaction volume have all exceeded $10 billion. Machi spent $4.11 billion but lost $4.16 million. After receiving tokens and rewards, the net loss was only $74700 On one hand, trading volume can be incentivized to be flushed out, and on the other hand, leverage can be turned into lottery tickets In the previous round of GMX's GLP, LP was also the counterparty, earning the money lost by traders. However, at that time, it had to be packaged as "decentralized sustainability". This round even saved packaging and was called YOLO directly So the question arises: Is the 1000x YOLO the endgame of Perp DEX's casino ization, or is it another signal of leverage peaking
+4
Mentioned
Share To

Timeline

HotFlash

APP

X

Telegram

Facebook

Reddit

CopyLink

Hot Reads