US CFTC Proposes to Bring Prediction Market Event Contracts Under Swap Regulation, Seeks Federal Regulatory Authority

According to Lookonchain’s monitoring, three wallets likely belonging to the same crypto whale transferred 2,903 ETH to Coinbase after remaining dormant for over 9 years. Valued at current prices, the transfer is worth approximately $7.22 million. Data shows that when these wallets received the ETH nine years ago, ETH was priced at just $9.9. The asset now holds an unrealized profit of around $7.19 million, with a return on investment (ROI) of roughly 250 times.

Beating AI Express News, according to a Bloomberg report, OpenAI and Anthropic have significant differences in their revenue accounting standards, making the annualized revenue figures released by the two AI giants difficult to compare directly, and intensifying investors’ concerns about the growth prospects of the AI industry. The report notes that Anthropic includes total sales generated through cloud service partners such as Amazon in its revenue, while OpenAI only recognizes the net revenue share it receives from partners like Microsoft. Since neither company has published standardized financial statements, investors currently struggle to accurately assess their actual business scales. Annualized revenue (ARR) also has limitations. Sources familiar with the matter revealed that OpenAI expects its full-year 2026 actual revenue to be around $35 billion, roughly half of its annualized revenue target of $700 billion by the end of the year. Meanwhile, documents reviewed by Bloomberg show that Anthropic’s actual 2025 revenue was approximately $4.6 billion, while its then-reported annualized revenue exceeded $9 billion. Analysts point out that annualized revenue is typically extrapolated from short-term performance or contract values, and cannot be equated to actual full-year recognized revenue. Against the backdrop of high valuations for AI giants and tech stocks trading at elevated levels, uncertainties surrounding revenue growth rates and accounting standards may further impact the market’s judgment of AI investment returns.

3 wallets (possibly belonging to the same whale) deposited 2,903 $ETH ($7.22M) into #Coinbase after over 9 years of inactivity. When these wallets received the $ETH 9 years ago, the price was only $9.9. They are now sitting on a $7.19M profit, with a 250x return.

Per monitoring by Ai Yi, F2Pool co-founder Wang Chun has recently shown greater optimism toward ETH’s trend. Yesterday, he swapped 235.51 WBTC for 7,848.53 ETH, a transaction worth roughly $19 million, executed at an ETH/BTC exchange rate of approximately 0.03. Over the past seven hours, he withdrew 5,173.41 ETH from Binance, valued at around $12.86 million, with an average withdrawal price of $2,485.89. Data shows that since October, the ETH/BTC exchange rate has continued to decline, falling from 0.03212 to 0.03013.

Beating AI Express: OpenAI Product and Platform Lead Thibault Sottiaux has announced Day 5 of the company’s 28-day product sprint. The update includes two features for ChatGPT: Dot and Codex. Dot is a personal AI assistant capable of continuously tracking tasks. Users eligible to access Dot can now create it directly in the iOS and Android versions of ChatGPT and converse with it on mobile devices, a change from the previous requirement of creating it on desktop. Additionally, OpenAI has enhanced task alignment between Dot and Codex: Dot can reference prior ChatGPT conversations, Codex tasks, and automation records to more accurately determine whether to continue an existing task or start a new one. For desktop Codex, the company is launching a test of Composer predictions: after Codex generates a reply, the input box may suggest the next message based on the current conversation. Users can fill in the suggestion by pressing the Tab key, then modify it before sending. Currently, this test is only available to ChatGPT Pro users aged 18 and older. The predictions themselves do not count toward Codex usage limits, though sending messages still counts toward regular usage quotas.

Beating AI News Flash: Microsoft Unveils Microsoft-Decision-1, AI Model for Rapid Agent Judgments Microsoft has released Microsoft-Decision-1, a model purpose-built to help AI agents make fast decisions. Post-trained on Qwen3.5-9B, it only scores pre-defined options and returns probabilities for each choice. Use cases include routing user requests to the appropriate model, or determining if an agent should proceed, retry, or escalate to human support. Microsoft states Decision-1 ranks first in accuracy across 36 evaluations and nearly 150,000 test questions. It is 4.5 times faster than the second-place Quyet-1.0-Large and roughly 35 times faster than GPT-6 Sol. In internal trials, Microsoft’s Xbox Research team used it to categorize over 10,000 pieces of player feedback. The company adds that Decision-1’s judgment quality is on par with GPT-6 Sol, while being more than 14 times faster and costing approximately 1/200th of the latter’s expense. The Copilot team also leverages it to score AI-generated responses. Decision-1 is now available on Microsoft Foundry and OpenRouter, with Vercel AI Gateway also integrated. It is priced at $0.042 per million input tokens, with output tokens free of charge. Microsoft has not yet announced plans to release Decision-1’s model weights.