MicroStrategy bought another 2,530 $BTC($243M) at an average price of $95,972 last week!

Dongcha Beating AI News Flash, as cited by media outlet Dujia, reports that Tencent’s Yuanbao is developing a personal AI agent named "Yuanbao Assistant", with plans to launch a standalone app. The new product will operate in its own space with separate permissions from the main Yuanbao app, focusing on long-term memory and continuous task execution. Users can assign long-term follow-up tasks to the AI without needing to restart conversations each time. Multiple business departments within Tencent are also working on personal AI agent-related features, including its social business unit. Recently, Ant Group’s Lingguang has started internal testing of long-term task hosting functions, while ByteDance and Alibaba are also advancing similar products.

According to official data, HyperEVM-based synthetic perpetual protocol Papertrade has been live for 36 hours since its Saturday evening launch. After an initial boom, trading has cooled sharply, with trading volume and open interest (OI) plummeting. The platform’s cumulative notional trading volume exceeds $770 billion, while its total value locked (TVL) stands at $62.6 million, down 37% in 24 hours. Open interest positions hit $510 million, a roughly 85% drop from $3.3 billion 24 hours prior. The total margin for all existing positions is now just $6.4 million, down around 70% from $20.29 million 24 hours ago. Papertrade is a fully on-chain synthetic perpetual trading platform deployed on HyperEVM, fetching prices from the best bid/offer (BBO) mid-price of Hyperliquid’s order book. Users trade directly against the protocol’s liquidity providers (LPs) instead of via traditional order book matching. The protocol features high leverage, zero slippage, and no funding fees. Its native token PAPER has an initial supply of zero, with no pre-mined tokens, team allocations, or VC shares. The only source of minted PAPER is user losses or liquidations. During its early launch phase, the token is not freely transferable and is primarily used for staking.

The newly created wallet 0x1389 withdrew another 600,007 $UNI ($4.61M) from #Binance an hour ago.

Based co-founder Edison released a statement on the recent card data leak, noting that the team’s compliance and operations departments have begun contacting and responding to affected users, notified the Singapore Personal Data Protection Commission (PDPC) and relevant regulatory authorities, filed a police report, and are currently cooperating with law enforcement in the investigation. Edison said the card business remains operational, and Based’s transaction wallets and app have not been impacted. The wallet uses a self-custody model, meaning the platform does not hold users’ private keys; its social login function is powered by Privy, and users can still conduct transactions on Based. Addressing external discussions about an alleged "admin backend leak five months ago," Edison clarified that this backend is unrelated to the current card data breach and contains no user data. He explained that the backend was an early development version of a hosted OpenClaw product the team tested in March 2026, intended only for small internal testing; the team later decided not to launch it and halted development. Edison added that the team’s current priorities include providing ongoing support to affected users, further strengthening security controls and monitoring systems, and continuing to build a higher-quality transaction experience.

According to Lookonchain's monitoring, a newly created address withdrew 600,007 UNI tokens from Binance one hour ago, valued at approximately $4.61 million.

Beating AI News Brief: OpenAI Product and Platform Lead Thibault Sottiaux has announced Day 7 of the company’s 28-day product sprint. GPT-6.1 Sol’s default context window in Codex and ChatGPT Work has been expanded from 272,000 tokens to 500,000 tokens, an approximately 80% increase. The larger window reduces instances of automatic context compression during long tasks, enabling the model to retain more details from prior conversations and task inputs. Sottiaux noted the new default setting was specifically tested, and users can still adjust the window size up or down, with a maximum limit of 1 million tokens. Notably, while larger context windows were previously only available via manual configuration, this update makes the expanded window the new default. Additionally, OpenAI will no longer charge long context surcharges for GPT-6.1 Sol users on ChatGPT subscription plans.