Big Whales' Movements

Binance’s He Yi says the exchange is working with U.S. companies and law enforcement to recover stolen assets

Binance co-founder He Yi said the exchange is coordinating with U.S. companies to freeze and recover related assets. She said U.S. law enforcement has been highly efficient in its involvement. He Yi also said that in cases drawing global attention, any information released by law enforcement agencies is subject to strict verification and backed by evidence. The remarks were reported by Odaily in a 7x24 news update. No further details on the assets or the case were disclosed in the brief.

XPL

15 hours agoNews Editor

XPL Crash Leaves Hyperliquid’s HLP Vault With About $556,000 in Losses

On-chain data shows that a sharp drop in XPL triggered a chain liquidation of long positions and left Hyperliquid’s HLP vault with an estimated loss of about $556,000. Before the sell-off, 16 wallets funded through centralized exchanges had been building XPL long exposure since Oct. 29, accumulating roughly 269 million XPL with a notional exposure of about $23 million. Of those, funds for 14 wallets were transferred through Binance around two months earlier, while 12 wallets withdrew about 2.02 million USDC before the price decline to reduce margin exposure. As XPL fell quickly from $0.086431 to $0.069429, the liquidation process forced the HLP vault to take over about 160.39 million XPL in long positions at an average entry cost of $0.07592, with a notional value of about $12.18 million. The vault later closed the position at $0.0725, resulting in the reported loss.

Deribit

16 hours agoNews Editor

Deribit settles today’s BTC and ETH options at $77,785 and $2,266

Deribit data shows that BTC options expiring today were settled at $77,785, which was 0.3% below the max pain point of $78,000. ETH options expiring today were settled at $2,266, or 0.3% above the max pain point of $2,260. The update was cited by Techub News and focused on the settlement prices for the day’s expiring contracts. No additional market context or trading figures were provided in the source. The figures indicate where the contracts were settled relative to their respective max pain levels, with BTC finishing slightly below that reference point and ETH slightly above it.

Stanford HAI

15 hours agoNews Editor

Stanford HAI says enterprise AI investment hit $581.7 billion in 2025

Stanford HAI’s 2026 AI Index Report said total enterprise investment in AI reached $581.7 billion in 2025, up 130% from a year earlier. The report said companies sharply increased spending on AI to improve efficiency and profitability. It also projected that hyperscalers including Microsoft, Amazon, Google and Meta will spend a combined $700 billion on AI infrastructure this year. As third-quarter earnings season gets underway, some companies have already started disclosing specific financial gains tied to AI, according to CNBC. The figures point to continued heavy capital deployment by large technology companies as AI spending moves from broad strategic commitment to measurable business results in corporate reporting.

15 hours agoNews Editor

CZ says Bitcoin avoided deeper losses largely because of U.S. law enforcement help

Changpeng Zhao, widely known as CZ, said Bitcoin did not fall further largely because U.S. law enforcement agencies played a key role after the Binance hack. According to his account, the Federal Bureau of Investigation, the U.S. Secret Service and the Internal Revenue Service helped track and freeze funds tied to the incident. CZ said a single post could not fully express his gratitude and added that many crypto users may not realize how much those agencies contributed to preventing additional downside in Bitcoin. Earlier, the FBI said North Korea was behind the theft of more than $6 billion from Binance. The remarks frame the agencies not only as investigators in the case, but also as an important stabilizing force during a period of market stress tied to the hack.

Uphold

15 hours agoNews Editor

Uphold RLUSD Earn and Ethena Pay Earn+ Compared on Lending and Leverage

The Defiant published a comparison of Uphold’s RLUSD Earn and Ethena’s Pay Earn+, framing the two products through a simple $10,000 illustration. The piece centers on how the strategies differ in lending mechanics and leverage exposure rather than treating them as interchangeable stablecoin yield products. Its main takeaway is straightforward: a leveraged stablecoin strategy can slip into negative returns if borrowing costs rise, even when there is no depeg. The article also tags Ethena Labs and places the discussion in the borrowing and lending category, signaling that funding costs and structure sit at the heart of the comparison. Based on the available source text, the report does not provide additional figures beyond the equal-$10,000 example, but it clearly highlights the risk that higher financing expenses can erode or reverse yield in leveraged setups.

100

OpenAI

21 hours agoNews Editor

Sam Altman says putting DRAM in personal PCs is wasteful and argues memory should be concentrated in AI data centers

OpenAI CEO Sam Altman said in a long interview with YouTube channel Cheeky Pint that scarce DRAM is being used inefficiently when it sits idle in personal computers most of the time. He argued that, from a global resource-allocation perspective, memory should be concentrated in cloud data centers where utilization can be much higher. The remarks touched a nerve because OpenAI’s business is built around closed models running in large-scale data centers, and critics see the argument as closely aligned with that model. Altman also discussed a wider set of issues. He said AI systems are getting much better at finding software vulnerabilities and warned that large-scale AI code scanning could reveal far more severe flaws than people expect. He argued patching alone will not be enough and said routers, operating systems, and application layers may need to be redesigned for a world where vulnerabilities become “more intelligent.” On AI safety, he said sandboxing only works to a point, monitoring can help, but alignment remains essential. The interview also covered OpenAI’s work on custom chips, data centers, robotics, and a “self-replicating data center” concept, as well as the economics of memory supply, AI agents, and pricing. Altman said OpenAI once came close to charging by task instead of tokens, but customers pushed back because they wanted a clearer sense of what they were paying for.

Tsinghua University

15 hours agoNews Editor

Tsinghua team led by Andrew Yao unveils AWM, scoring 99.9% on ARC-AGI-3

A research team from Tsinghua University led by Andrew Yao and Yuan Yang has introduced the Algebraic World Model, or AWM, a system designed to help AI move from observation and experimentation to mathematical abstraction of how the world works. According to the team, AWM achieved a 99.9% score on the ARC-AGI-3 benchmark while using an average number of actions equal to 43% of the human baseline. The framework follows a three-step process: it first explores phenomena and writes hypotheses in Python code, then tests those hypotheses against new observations to extract general rules, and finally formalizes the rules in Lean 4 for mathematical proof. For unsolvable tasks, the Opus 5.5-based version of AWM reportedly needed only two attempts to prove that no solution existed and to identify the number of steps required to win. The research team includes co-first authors Pan Guancheng of Zhejiang University and Tan Qixin of Tsinghua’s Yao Class, with Yuan Yang and Turing Award winner Andrew Yao serving as advisors. The paper has been posted on arXiv, and the project homepage has also been released.