US after-hours trading rebounds for semiconductor, optical communication, and storage-related concepts; LITE's CEO states that all optical device production capacity will have been fully sold out by early 2029.

According to an official announcement, Binance will add watchlist tags to BICO and CVC on October 9, 2026. Tokens with watchlist tags exhibit higher volatility and risk compared to other listed tokens. Trading these tagged tokens is risky, as they no longer meet the platform’s listing standards and are subject to potential delisting.

Beating AI Express: OpenAI’s research chief has issued a statement responding to an open letter from three former employees—Jasmine, Mikita, and Tomek. The company said that following a full investigation, the three violated explicit sensitive information handling policies, and the internal probe uncovered serious breaches of trust that went beyond what was disclosed in the open letter. As a result, OpenAI terminated their employment last week and is standing by this decision. OpenAI emphasized that the decision was not targeted at employees raising safety concerns or expressing opinions publicly. The company added that it holds intense, critical safety and research discussions internally every day, encourages staff to raise questions, is tolerant of good-faith mistakes, and will not fire employees for voicing concerns. Regarding the safety issues covered in the open letter, OpenAI stated it is finalizing a contract with a third-party security assessment firm, expects to release details in the coming weeks, and commits to integrating external assessors into its safety work. The company also acknowledged that maintaining the monitorability of frontier models requires joint efforts from the entire industry, noting that this area has long been a core research focus for OpenAI and it will continue to invest significant resources in this field.

Strategy founder and executive chairman Michael Saylor told Binance in an interview that the period from now through 2035 will be a "gold rush" era for digital assets. The crypto economy, still smaller than other global asset markets, is in a high-growth phase. He noted that 99% of all Bitcoin will have been mined by 2035, which he cited as one key reason the industry remains in its early stages. Saylor positions Bitcoin as "digital capital," identifying four major drivers for its adoption: ETFs and wrapped Bitcoin products, treasury companies, digital credit, and bank-collateralized credit. He emphasized that Strategy has no fixed Bitcoin holding target: "We accumulate Bitcoin as we raise capital." The firm’s holdings have risen to approximately 848,000 BTC. He explained that digital credit products like STRC are designed to offer volatility and cash flow characteristics different from direct Bitcoin holdings, catering to investors who cannot tolerate Bitcoin’s price swings. STRC is a perpetual preferred stock issued by Strategy that pays cash dividends. Additionally, he highlighted focus on Bitcoin’s 200-week moving average and its price premium or discount relative to this metric, adding that funds previously flowing into AI and large private equity deals have started returning to the crypto market.

Crypto trader Ansem said he views the current correction as a potential opportunity to add to high-conviction positions. Bitcoin recently tested its year-to-date opening price of around $87,500 before pulling back to near $81,700. He is closely watching the key support level of $75,000 below; if Bitcoin reclaims the $87,500 mark, it will reinforce the bullish outlook. Ansem noted that interest rates and geopolitical tensions remain the largest macro risks at present. As long as the market structure holds up, he prefers to gradually build positions amid volatility.

Beating AI News Flash: Shenzhen-based diffusion language model firm DiffuSpace has closed two consecutive financing rounds totaling nearly 500 million yuan. The rounds were jointly led by Matrix Partners China, Shunwei Capital, and Legend Capital, with follow-on investments from Huawei Harbin Investment, Horizon Robotics, and others. Multiple media outlets note this is the largest financing round in the global diffusion language model space. DiffuSpace was founded in May this year by The University of Hong Kong professor Kong Lingpeng and his PhD students Gong Shansan and Ye Jiacheng. The team previously co-developed Dream 7B with Huawei Noah’s Ark Lab and has open-sourced the model’s weights. It uses diffusion large language model (dLLM) technology. Unlike mainstream models like GPT and Claude that generate text sequentially, diffusion models can process content across multiple positions simultaneously before gradually completing and refining it. DiffuSpace is currently training a new 30-billion-parameter model, with plans to release and open-source it soon. The company will focus on advancing code agents and local deployment on devices including smartphones, automobiles, and robots.

Beating AI Express: World model company Odyssey has officially launched Odyssey-3, with its Flash version now available for free trial. The model can generate explorable worlds in real time from text, with visuals updating dynamically as users move or rotate the camera. The team also showcased experiments applying the same model to robotic arms, humanoid robots, vehicles, and game characters. Odyssey-3 first learns object movement and action outcomes from internet videos, game play recordings, and physical simulations. When deployed for robots or vehicles, the main pre-trained model parameters remain unchanged; the team only needs to train additional small control modules to translate the model’s learned knowledge into operational instructions. While different machines still require separate adaptation, full retraining of the entire model is unnecessary. Odyssey trained a dedicated control module using just 20 hours of simulated driving data, enabling a car to operate on real roads in India. In tests, the simulated data training scheme delivered approximately 77% of the driving distance between two safety officer takeovers compared to the real driving video training scheme. For the humanoid robot developed in partnership with Flexion, after training on tens of hours of remote control demonstration data, it can complete tasks even under varying lighting conditions. In game experiments, the control module learned from roughly 2 hours of Grand Theft Auto V (GTA V) gameplay footage, and without additional training, it enabled characters to ride horses and move in Red Dead Redemption 2. In the Physics-IQ Verified physical video prediction test, Odyssey-3 Pro scored 66.1 points, surpassing FLUX 3 [large]’s 64.35 points, with both results being the best from 8 trials. Odyssey also released WorldMark evaluation results, where the model secured first place in three out of four interaction scenarios.