Markets are pricing in four more interest rate hikes from the Federal Reserve through June next year, with Bitcoin briefly dipping below $83,000.

Binance announced in an official statement that it will gradually add Hyperliquid (HYPE) to its services including Simple Earn, Buy Crypto, Convert, VIP Loan, and Margin starting September 24. HYPE’s flexible product will launch on Simple Earn at 19:00 UTC+8 on September 24, with subscriptions opening immediately. Within one hour of HYPE’s spot trading going live, users can purchase HYPE via payment methods such as bank cards, Google Pay, and Apple Pay, and trade HYPE with assets like BTC and USDT on Binance Convert with zero fees. Additionally, HYPE will become an eligible borrow asset for VIP Loan within one hour of its spot listing. Binance’s margin trading will add HYPE as a borrow asset at 19:00 UTC+8 on September 24, and launch HYPE/USDT and HYPE/USDC cross-margin and isolated margin trading pairs; portfolio margin will also support HYPE and the above-mentioned trading pairs simultaneously. Binance reminded that newly listed tokens usually have high volatility, so users should pay attention to risk management.

Beating AI Express News: Apple has open-sourced LensVLM-9B, a vision-language model purpose-built for processing long documents, trained on Qwen3.5-9B-Base. When handling long documents, the model first compresses the entire document into low-resolution pages for a quick scan to locate relevant pages, then reads the original text or high-definition images on those pages—eliminating the need to load the full document into the model’s context window. Per the paper’s tests: For a 100-page document, reading the full text directly requires 51,273 tokens, while LensVLM uses just 8,090 tokens. The corresponding KV cache drops from approximately 1.6GB to 253MB, an 84.2% reduction, with no significant accuracy loss after compression. Among 7 document question-and-answer tasks, direct full-text reading yields an average accuracy of 72.4%, while LensVLM hits 68.9% at roughly 4.3x compression. Compared to the approach of converting text to images for model recognition (at around 5x compression), LensVLM lifts accuracy from 31.3% to 68.9%. That said, the model is currently slower: After locating the relevant page, LensVLM must call an additional tool to read the original text, requiring two consecutive inference runs. The paper notes each answer takes approximately 17 seconds; without the compression scheme, directly inputting the full document as text into Qwen3.5-9B takes about 8 seconds to generate a response.

Bitget Onchain, its on-chain trading service, has now integrated Robinhood Chain, enabling users to directly trade all assets in the Robinhood ecosystem using the stablecoin U (United Stables) from their spot wallets. This integration further expands Bitget UEX’s panoramic ecosystem layout, delivering a more efficient on-chain trading experience for users. As of now, Bitget Onchain has integrated six major networks: Robinhood, MORPH, ETH, SOL, BSC, and Base. Users can directly access millions of on-chain assets and US stock trades via their spot wallets, and leverage auxiliary tools including AI smart money copy trading and on-chain limit orders to achieve intelligent analysis and trade execution.

Insight Beating AI Flash News: Personal AI assistants have been popping up en masse lately, and some in Silicon Valley are already poking fun at these products. Tech video creator Max Musgrave has developed a fake product called Sidehoe, billed as "the first AI assistant to help you manage your dating pool". It reads iMessage chats, remembers the names, birthdays, and preferences of each date, alerts you to who you’ve been ghosting for 9 days, checks for conflicting appointments, misnamed messages, and whether the same flirty line has been sent to multiple people. Sidehoe has copied nearly all standard features of personal AI assistants: it stores information, proactively reminds users, drafts responses, and claims all data is processed locally. To avoid being detected by partners, notifications are disguised as delivery alerts, and the app is hidden behind a seemingly functional calculator. It even offers a $9.99/month Pro plan, plus a "Monogamy" plan for managing only one person, which notes "No app required". Only at the bottom of the page does it admit that Sidehoe is a parody, not an actual product. Musgrave runs Strive Haus, a San Francisco-based creative studio that specializes in producing product launch videos and brand content for tech companies. Familiar with these launch routines, Sidehoe’s product demo, feature naming, and subscription page are all highly realistic.

Flash News from Dongcha Beating AI: Today, a personal AI assistant founded by Qi Junyuan—founder of Teambition and former head of Doubao PC—has officially launched in China. Today was previously in open beta overseas, and this marks its first time offering downloads and registrations to Chinese users. The basic version is permanently free; new users receive 100 million Tokens and a 1-month Pro subscription. It currently supports Mac, Windows, Linux, iOS, and Android. Today continuously accesses users’ authorized emails, calendars, notes, and documents, and retains key contacts, preferences, commitments, and pending tasks. When tasks have new updates, it proactively reminds users and leverages cloud computing to handle operations on web pages, files, and other applications. In a prior test by PingWest, users asked Today to schedule an interview. It first located contacts via Gmail, checked the calendar, recommended a time, drafted an email, and sent it directly after user confirmation. The next day, after the recipient replied, the journalist issued no further instructions—Today independently picked up the new email and the previous interview task, and continued to provide next steps. Accessing emails, checking calendars, and sending messages on behalf of users are not new capabilities. What differentiates Today is its ability to keep tasks in the system after a conversation ends, and follow up when triggers like email replies, schedule changes, or other conditions arise. However, this also poses challenges: the more it remembers and the more proactive it is, the higher the requirements for memory accuracy, permission control, and privacy protection.

Swiss fintech firm Colb announced on September 24 the launch of its digital finance AI agent Colbee, which is now live on WhatsApp. Designed to simplify access to DeFi and on-chain financial products through conversational interactions, Colbee allows users to explore and access relevant services via natural language without directly operating wallets, networks, or decentralized exchanges. Colb stated that Colbee previously served primarily accredited investors, having executed over $40 million in crypto and tokenized asset trades in line with client instructions and signatures. Currently, the AI agent supports 10,163 on-chain operations across Ethereum, BNB Chain, and Polygon, including 63 product functions, 3,368 swap paths, and 6,732 cross-chain paths, with integrations with infrastructure such as PancakeSwap. On the security front, Colbee does not hold user private keys or funds, nor does it make autonomous investment decisions. Each transaction generates an unsigned request based on user instructions, which undergoes checks for balance, authorization, and product limits plus on-chain simulation before users review and sign the request via their own wallets. Some tokenized financial products still require accreditation, identity verification, and compliance with regional restrictions. Looking ahead, Colb plans to expand Colbee to a wider range of on-chain financial services, with a target to grow its platform’s total value locked (TVL) to $1 billion by 2027. As of now, Colb’s colb.finance holds a TVL of over $108 million.