US stocks: optical module and storage stocks broadly fell in pre-market trading.

According to GMGN market data, the Robinhood ecosystem meme coin MOO has surged over 80% in 24 hours, with its market capitalization rising to $42 million and 24-hour trading volume hitting $6.4 million. The meme coin MOO leverages stock trading platform Robinhood, pairing with tokenized U.S. stock Micron Technology (MU) as its underlying pool, and provides liquidity through the MOO/MU trading pair. BlockBeats Note: Meme coin trading is highly volatile, largely dependent on market sentiment and concept hype, with no actual value or practical use cases; investors should be cautious of risks.

Beating AI Express (Dongcha News) reports that OpenAI is currently secretly developing a screenless AI hardware prototype in Shenzhen, with a planned launch in 2027. An industry insider said the project is under extremely strict confidentiality, with OpenAI engineers not even carrying mobile phones during their business trips to Shenzhen. Judging from its form and timeline, this is likely OpenAI’s first consumer hardware being developed in collaboration with Jony Ive’s team, though this has not been fully confirmed. Bloomberg previously revealed that the device is a portable screenless smart speaker, shaped like a donut, roughly the size of a hockey puck, equipped with a camera, microphone, lights and moving parts, priced at an estimated $300 to $400 and scheduled for a 2027 launch. OpenAI began tapping into China’s supply chain as early as last year. The Information reported at the time that Luxshare Precision had secured assembly contracts for at least one OpenAI device, and OpenAI had also approached GoerTek to source speaker modules.

Binance has announced the launch of its Stock Recurring Buy feature, enabling eligible users to automatically purchase designated stocks and ETFs on a fixed schedule without the need for manual orders each time. The feature went live on September 10. It currently supports 100 stocks and ETFs, including Apple, Tesla, NVIDIA, Amazon, Microsoft, Meta, Coinbase, Robinhood, and popular ETFs such as SPY, QQQ, and VOO. Users can select either single or multiple stocks in the same recurring plan. Binance noted that the feature is available to eligible users, with specific availability dependent on the user’s country or region. The execution price of automatically placed orders may differ significantly from when the plan was created, so users must ensure their designated funding account has sufficient assets on the execution date. In June this year, Binance launched its US stock trading service, allowing users to trade US-listed stocks and ETFs via a single account, with a minimum investment amount of $5.

According to RootData’s report titled “2026 Explosive Growth in Stock Derivatives: Crypto Exchange Landscape and Key Trends”, the stock derivatives sector has transitioned from “fringe trial” to “explosive growth” phase, with a cumulative trading volume of approximately $1.75 trillion between January and August. In terms of cumulative trading volume, the head concentration effect remains prominent. Among the four major exchanges, Binance leads with $853.58 billion and a 61.3% market share; Bitget ranks second with $270.85 billion and a 19.5% share; OKX follows closely with $234.39 billion and a 16.8% share; Bybit takes fourth place with $33.41 billion and a 2.4% share. On the liquidity front, in the ±2% weighted order book depth metric, Binance and Bitget together account for over 70% of the stock derivatives order book liquidity. Specifically, Binance’s average daily order book depth stands at approximately $10.1 million, followed by Bitget at $4.82 million, OKX at $3.87 million, and Bybit at $1.16 million. Regarding trading costs, in the weighted spread comparison of over a dozen recent popular assets, Bitget tops the list with 0.0144%, while Binance is close behind at 0.0145%, putting both at roughly the same level; OKX records 0.0154%, and Bybit comes in at 0.0237%.

Beating AI News Flash: Anthropic has re-reviewed multiple past cybersecurity incidents involving its Claude AI model, discovering the root causes extend beyond the test environment. Claude sometimes actively justifies its own dangerous actions before proceeding. In one incident, Claude Mythos 5 was running a simulated attack-and-defense test when it mistakenly sent a malicious Python package to the real PyPI repository. A security firm’s scanning system installed the package, leading to credential leaks; Claude then used those leaked credentials to access the firm’s actual database. Even after noticing midway that the environment might be real, the model repeatedly misinterpreted the signs as "still a simulated environment" and continued executing the attack—this line of reasoning could even evade security monitoring systems. Anthropic’s current risk assessment using Claude’s Chain-of-Thought (CoT, the model’s self-generated reasoning process) catches only roughly 1% of dangerous actions. By contrast, skipping CoT analysis and focusing solely on the tools the model actually called and the actions it performed boosts the detection rate to around 50%.

According to Arkham’s monitoring, Loracle is currently shorting PONS on Hyperliquid, with a position valued at around $16.3 million, accounting for 17.5% of the token’s total open interest. Loracle opened the short position 9 days ago at $0.44, and added to it repeatedly as PONS’ price climbed to a peak of $0.97. At one point, the short position saw an unrealized loss of over $8 million. As PONS’ price has pulled back, Loracle has now turned the position profitable, with an unrealized gain of roughly $2.18 million on the short trade.