U.S. spot Bitcoin ETFs have attracted inflows for 7 consecutive days, with their year-to-date net outflow gap narrowed by more than half.

Per monitoring by on-chain analytics platform Lookonchain, trader 0xD81a—who previously incurred roughly $12 million in cumulative losses from buying Ethereum at high prices and selling at low—has re-entered the market after nearly six months of inactivity. The address spent approximately $5.33 million to purchase 2,165 ETH at an average price of around $2,463. Notably, the trader had previously cut their ETH holdings at an average price of about $2,452 in their last trade, only to repurchase the asset at a higher price now, resulting in a "sell low, buy high" operation.

Beating AI Express News: The identity guessing game around the mysterious model Ox Alpha has clearly tilted toward Zhipu AI. On Polymarket, the probability assigned to Z.ai currently exceeds 90%, while second-place Google holds only around 5%. Ox Alpha is an anonymous inference model that recently emerged on OpenRouter, focusing on programming and long-duration agent tasks. The most concrete clues come from the backend: community developers intentionally sent incorrect parameters to Ox Alpha’s OpenCode interface, and the server returned com.wd.paas.api.domain.v4.chat.ChatCompletionRequest. The "paas/v4/chat" segment in this response matches Z.ai’s official API path exactly. When invalid "role" parameters were sent, Ox Alpha returned error code 1214: Incorrect role information, which aligns with Z.ai’s self-hosted GLM model. When the same GLM weights were hosted on DeepInfra, the error format changed. The model’s own fingerprint also matches. A public forensics test ran over 600 requests, using approximately 13.5 million input tokens, with all 44 tokenizer tests matching the GLM-5 generation.

Approximately 81,700 Bitcoin options will expire on Deribit at 16:00 Beijing time this Friday, with a notional value of around $6.44 billion, which could further amplify market volatility. Of these, about 44,600 are call options and 37,100 are put options, with a put/call ratio of 0.83, indicating an overall bullish bias. The most concentrated strike prices for call options are $75,000 and $80,000, corresponding to open interest notional values of roughly $236 million and $157 million respectively. Deribit’s Chief Risk Officer noted that Bitcoin surged from around $62,000 to $80,000 over the past week, pushing a large number of call options into in-the-money territory and putting market makers under greater hedging pressure. Currently, options within 5% of the current price have a notional value exceeding $500 million. The approaching expiry could trigger more frequent gamma hedging, leading to a "pinning" effect on prices near key strike prices, or accelerated volatility after breaking through critical levels.

At 20:30 Beijing time tonight, the U.S. Department of Commerce will release the July Personal Consumption Expenditures (PCE) Price Index. Market forecasts show that July’s headline PCE will rise 0.1% month-over-month, with its year-on-year growth rate falling from 3.7% in June to 3.6%. Core PCE is expected to climb from 0.1% to 0.2% month-over-month, while its year-on-year rate holds at 3.3% — marking the 65th consecutive month above the Federal Reserve’s 2% inflation target. Notably, price increases in the AI industry chain, higher stock market valuations driving portfolio management fees, and rising energy costs due to Middle East tensions could all act as potential drivers of core inflation. Goldman Sachs estimates the stock market valuation factor alone may contribute roughly 0.11 percentage points to July’s core PCE month-over-month growth. What draws even greater market attention is that the U.S. Bureau of Economic Analysis (BEA) plans to fully revise its PCE calculation methodology by the end of September. Going forward, adjustments may be made to price calculations for categories including computer hardware, stock portfolio management, and legal services, with possible historical data revisions that could complicate the interpretation of inflation figures. Currently, divisions over the Federal Reserve’s September policy path are intensifying. CME Group’s FedWatch Tool shows a 59.9% probability the Fed will hold interest rates steady in September, while the chance of a 25 basis point hike has risen to 40.1%. The market is also awaiting this week’s Jackson Hole Economic Symposium to seek the latest signals from Fed Chair Powell on inflation and future interest rate policies.

Insight Beating AI Flash News: Sources familiar with the matter disclosed that Moonshot AI is in negotiations with Microsoft, Amazon, and Google regarding revenue-sharing agreements, under which the three major U.S. cloud computing giants would host its Kimi K3 model. Insiders said Moonshot AI is seeking up to a 30% share of revenue from Kimi K3-related services on Microsoft Azure, AWS, and Google Cloud, with terms largely aligned with those it provides to large clients using its open-weight models. Currently, Moonshot AI has already signed similar deals with some smaller cloud platforms. If the negotiations are concluded, the agreement could mark the first major model revenue-sharing collaboration between a Chinese AI company and a leading U.S. cloud computing firm. Meanwhile, Alibaba is also working to establish a revenue-sharing mechanism with key users of its new open-source AI model.

Beating AI News Brief: Jiyuan Ludong, founded by former head of Huawei's Pangu large model Wang Yunhe, has completed a new round of financing, with a valuation of several hundred million U.S. dollars and cumulative total financing exceeding tens of millions of U.S. dollars. The round was led by Honghui Fund. The company's core product is Routing Harness, which enables agents to automatically select and switch between different models during operation. Its OpenSquilla project has gained around 6,400 stars on GitHub; its multi-model API platform now has 54,000 users, with daily token calls exceeding 500 billion. In evaluations released by Jiyuan Ludong, a domestic multi-model combination outperformed Fable 5 on the DRACO research task, at approximately one-third of Fable 5's cost. Wang Yunhe's next plan is to train his own models, leveraging real tasks, model selection data and user feedback accumulated from the Harness, to advance from "helping agents pick models" to building Agent-Native Model capabilities.