Ayer

Please don’t use this! do not use the free bot to search over 85+M papers. Please ignore that you funded the research with your taxes, the authors paid $3,000 to publish it, and the peer reviewers worked for free. Please subscribe at $40 per article so Elsevier can keep its margins. Please do not bookmark sci-bot dot ru

Finance Progress. $CHIP is live on @RobinhoodApp.

Big moment coming for local AI. @nvidia x @apple

Global risk is being priced this weekend in real time on @tradexyz and @HyperliquidX

This image is insane!

AI-powered scientists are starting to take off! This paper introduces PHYSMASTER, an LLM-based agent designed to operate as an autonomous theoretical and computational physicist. The goal is to go from an AI co-scientist to an autonomous AI scientist in fundamental physics. PHYSMASTER uses Monte Carlo Tree Search for adaptive exploration, hierarchical agent collaboration for complex tasks, and LANDAU, a layered knowledge base preserving retrieved papers, curated prior knowledge, and validated methodology traces for reuse. Five case studies spanning from the cosmos to elementary particles demonstrate its capability: Two acceleration cases: PHYSMASTER compressed labor-intensive engineering work that typically takes a senior PhD 1-3 months into less than 6 hours. Two automation cases: Given human-specified hypotheses, PHYSMASTER executed end-to-end exploration loops in 1 day rather than unpredictable months. One autonomous discovery case: PHYSMASTER independently explored an open problem in semi-leptonic decays of charmed mesons, constructing the effective Hamiltonian and predicting decay amplitudes using SU(3) flavor symmetry. Paper: arxiv.org/abs/2512.19799 Learn to build effective AI agents in our academy: dair-ai.thinkific.com

USDAI is partnering with @PayPal to integrate $PYUSD as a settlement asset for AI infrastructure financing. To support adoption, PayPal and USDAI are launching a customer incentive program offering 4.5% on up to $1 billion in USDai deposits, expected to begin in Jan 2026.

A little over three years ago at BTC 17k. Good example of Central Bank analysis and forecasting ability.

ClusterMAX 2.0: The Industry Standard GPU Cloud Rating System 95% Coverage By Volume, 84 Providers Rated, 209 Providers Tracked, 140+ Customers Surveyed, 40,000 Words For Your Enjoyment newsletter.semianalysis.com/p/clustermax-2…

Debunking the FUD that Hyperliquid prioritizes protocol revenue over traders On 10/10, Hyperliquid ADLs net made users hundreds of millions of dollars by closing profitable short positions at favorable prices. If more positions had been backstop liquidated, HLP could have made hundreds of millions of dollars more in pnl, while being exposed to an irresponsible amount of risk. ADL passed on HLP's potential pnl to users while decreasing HLP's exposure, a win-win. As a reminder the ADL queue on Hyperliqid has always followed a similar formula to what most CEXs use, incorporating both leverage used and unrealized pnl on the open position. Finally, thanks to everyone for the feedback on ADL. Suggestions generally increase complexity, such as partially offsetting long and short positions in historically correlated assets. I don't know of other major venues that use more complex logic for the ADL queue. Simple formulas are more robust and understandable by users. Nonetheless, there is research being done on whether there can be substantial improvements that merit more complexity.

We’re all trying to compare different days, but there’s no reason for it. Yesterday was by far the biggest shock in crypto history - at a time when we were talking about the maturity of the space, bridging the gap between TradFi and crypto, and had started getting lazy with leverage. Tuesday 8pm UTC - flash crash breakdown. Stay tuned (and safe).

USDAI has structured a $500M non-recourse financing facility to fund GPU-backed AI infrastructure for @QumulusAI. This marks the largest onchain-based credit facility ever created for real-world compute.

Are you ready?

@sershokunin @hasufl gotta add the ability to whitelist addresses

@hasufl The design of being able to go direct from cex was an important one

It's funny that Hyperunit made it actually 10x easier to deposit ETH/BTC/SOL to Hyperliquid compared to USDC (which is the main asset!) Not exactly how you would expect these things to work

Excited to team up with @0xZergs and @_ConorMoore for USD.AI’s next chapter in building a new era for AI compute

@hedgedhog7 @cubantobacco only other solution was waiting for HLP to get liqd as nance and okex listed futures not good precedent but better that going down

generational entry on $HYPE inc as HLP takes on bad debt from a $JELLY liquidation

@0xLouisT 2 weeks of zero net inflows kek

Blackberry of crypto

@FloodCapital @FireblocksHQ @sershokunin I guess that's the excuse they used for having missed the entire move up

@0xAyer @FireblocksHQ @sershokunin Multiple funds I've spoken to were having trouble with custody previously.

Fresh $HYPE catalysts just dropped: @FireblocksHQ just announced support for HyperEVM which should unleash a ton of buying from funds who previously could not custody $HYPE (this is a huge deal many funds I've spoken to could not custody before this announcement) @sershokunin teasing stocks/ETFs on Hyperliquid spot markets via @unitxyz (cross margin on these would also be epic). Staking $HYPE for discounted fees is live, following the $BNB playbook. Hyperliquid.

@Galois_Capital has to be goblin town capital

I’m thinking of naming it either Reaper Capital, Garbage Collector LLC, Not My Bags Research, or Goblin Town Capital.