OpenAI conducted internal research on Ethereum testnet contract deployment, but an agent, facing insufficient gas, directly stole testnet tokens from an address with a publicly available mnemonic phrase.

Beating AI Express News reports that Zhang Yingya, head of Alibaba’s Tongyi Wanxiang (Wan series) video generation model, has recently left the company, with his next career move unconfirmed. Zhang joined Alibaba in 2015 via campus recruitment, having worked at the firm for around 11 years. Early in his tenure, he focused on visual recognition and image detection, later transitioning to diffusion models and video generation. In 2023, he took part in developing Alibaba’s early open-source text-to-video model ModelScopeT2V, and subsequently became a core R&D member of the Wan series. He is also credited in the Wan technical report released in 2025. The resignation follows Alibaba’s multimodal team restructuring: in July this year, the Tongyi Wanxiang team was merged into the Future Life Lab led by Vice President Zhang Di, aligning management with the HappyHorse video model and HappyOyster world model.

According to data from the Polymarket platform, trader dromedaryy spent $19,200 betting on the outcome of "Tesla Roadster flying for at least 1 second". The position is now valued at roughly $192,400, generating a profit of approximately $173,100, representing a return of 898.51%.

According to official announcements, X Layer and TapeOut have jointly launched the two-month "X Layer TapeOut Developer Acceleration Program", open to developers, creators, and on-chain communities. The initiative encourages participants to build on-chain applications on X Layer that meet real user needs, leveraging the TapeOut protocol and its ecosystem tools, to turn innovative ideas into practical, usable products. The program will feature activities including live streams, open AMAs, mentor Q&A sessions, and both online and offline Demo Days, providing participating teams with technical exchanges, product feedback, and resource matching. The official teams of X Layer and TapeOut will conduct joint reviews, evaluating projects across key dimensions such as product value, user experience, commercialization potential, and ecosystem technology applications. Outstanding projects will receive follow-up support in product, technical, and ecosystem collaboration.

Bitcoin treasury firm Capital B released its H1 2026 financial results, reporting revenue of €5.846 million, a 3.5% year-over-year rise; net loss reached approximately €69.5 million, compared to a net loss of €1.9 million in the same period last year. The loss was mainly driven by an additional impairment charge of around €67.5 million booked based on BTC prices as of June 30, which had no cash impact. As of September 30, the company’s strategic reserve held 3,538 BTC, a 25.3% increase from January, with a total acquisition cost of €310.6 million, averaging €87,805 per BTC. The announcement confirmed that as of October 15, the strategic reserve count remained at 3,538 BTC, with an extra 61 BTC held for operational use, which are not included in strategic reserve-related metrics. The company announced about €27.5 million in financing during H1, and another €31.7 million in financing between the end of the first half and the announcement date. As of September 30, the company’s available cash stood at €4.6 million, which, combined with the operational BTC, totaled roughly €9.1 million in value.

According to Bitget market data, Indonesia’s Jakarta Composite Index has dropped to 5193.22 points, with a decline of 11.4%, hitting its lowest level since early November 2020.

Beating AI News Flash: Nvidia has launched Nemotron 3.5 Super VL, releasing its model weights as open-source. The model has 120 billion total parameters, with only 12 billion activated per run, supporting a maximum 1 million-token context window. It combines text reasoning, visual understanding, and tool calling, allowing agents to directly read PDFs, analyze charts and videos, search the web, and complete tasks via tool calls—eliminating the need for repeated information transfer between text and vision models. The new model adopts a hybrid Mamba-Transformer architecture, improving inference efficiency through multi-token prediction and 4-bit quantization. In tests by Artificial Analysis, with a 100,000-token input, it achieved an output speed of roughly 630 tokens per second, about 5 times faster than Inkling Small, a model with comparable capability scores. Nvidia also noted that batch processing 10,000 SWE-bench Pro programming tasks using 4 B200 GPUs is up to 4 times faster than some comparable open-source models. Artificial Analysis gave Nemotron 3.5 Super VL a score of 26.5, ranking it first among U.S. open-source models, outperforming Nvidia’s earlier 550-billion-parameter Nemotron 3 Ultra (23 points) but still behind Xiaomi’s MiMo-V2.6-Pro (46 points). In Nvidia’s published SWE-bench Pro comparison, the new model’s pass rate stood at 51.78%, lower than Qwen3.8 Flash-Next’s 60.37%. Independent testing also exposed gaps between speed and reliability: Plotly evaluated the model on 43 data analysis questions, with the model answering 30 correctly, a median time of just 12 seconds per question, but missing all 3 data quality questions and 5 hidden pattern questions.