Jev is overtaken for the first time, and Liquid AI releases its decision-making model d1.

Former New York Governor and OKX board member Andrew Cuomo warned that if Democrats win control of Congress in the midterm elections, the crypto industry’s current regulatory progress could be overturned. He noted that interim rules drafted by regulators are far more fragile than legislation, stressing that “regulation is not legislation”. Earlier, the Senate on September 15 rejected the CLARITY Act—legislation intended to establish federal rules for the crypto market—by a vote of 49 in favor to 50 against, falling short of the 60-vote threshold needed for passage. On September 17, the U.S. Securities and Exchange Commission (SEC) granted a five-year exemption for tokenized stock trading. Cuomo anticipates the new Congress will conduct strict reviews of such agency rules, and expressed hope that relevant legislation could be passed by a lame-duck session or before January 1 next year.

Beating AI News Flash: Y Combinator (YC) startup InsForge has launched InstaCloud, announcing an $8 million seed financing round. A serverless cloud platform built for AI coding agents, InstaCloud enables tools like Claude Code and Cursor to directly deploy and run applications, offloading much of the previously manual DevOps work to agents. InstaCloud’s computing services auto-scale with traffic, shrinking to zero during idle periods—eliminating the need to pre-select machine specifications. Agents can create services, deploy code, view logs, and manage databases and storage via CLI or MCP. Its standout feature is "environment branching": Agents can replicate full stacks (Postgres, object storage, application containers) into isolated environments for code edits, database migrations, and testing. If issues arise, the branch can be deleted without impacting production, and critical operations can require manual approval. Co-founder and CEO Hang Huang directly stated InstaCloud’s goal to "disrupt AWS, GCP, and Azure". Currently, the platform is essentially an Agent-optimized iteration of cloud application platforms like Railway and Render, rather than a full replacement for the three major cloud providers.

Beating AI News Flash: Kuaishou has adjusted its AI organizational structure, establishing a cross-business unit virtual organization named "Enterprise AI Productivity" dedicated to the company’s internal AI office and R&D tools. The new organization will unify employees’ general Agents, internal corporate knowledge and data, as well as office and information systems, enabling Agents to directly call these resources to complete tasks. It is led by Wang Xu, head of the Intelligent Office Platform Department, and consists of the Intelligent Office Platform Department and the AI Productivity Center. Kuaishou has already deployed AI extensively internally. As of June this year, over 92% of its employees use self-developed AI Agents, with AI code contribution rate reaching 60% among R&D technical staff. Meanwhile, the company has adjusted the management division of Keling AI. Kuaishou co-founder, Chairman and CEO Cheng Yixiao will concurrently serve as head of the Community Science Line, while former head Yu Yue has been transferred to Director and CEO of Keling AI, focusing on the firm’s operational management going forward. Yu Yue (also known as Gai Kun) holds a PhD from Tsinghua University. He joined Alibaba in 2011 as part of the first batch of "Alibaba Stars", and before leaving Alibaba, he served as Technical Director of Precision Display Advertising at Alibaba Mom, holding a P10 level. After joining Kuaishou in 2020, he has long been in charge of algorithms and the Community Science Line, and led the R&D and commercialization of Keling AI. Keling AI’s revenue in the second quarter of this year exceeded 850 million yuan, with year-on-year growth of over 200%.

Gemini co-founder Tyler Winklevoss said the current market hype around privacy coin Zcash is reminiscent of the 2019 crypto cycle, when Bitcoin had just emerged from the 2018 bear market, with the market’s focus shifting to institutional infrastructure and custody. Zcash launched in 2016, uses zero-knowledge proofs to facilitate private transactions, has long faced regulatory scrutiny, and was delisted by some exchanges. Per CoinGecko data, ZEC’s price rose from approximately $511 in January 2026 to $1,335.51 on September 16. After receiving approval from the State of New York in 2018, Gemini became the first licensed trading platform to offer Zcash trading and custody services.

Beating AI Insight News Brief: Large model inference and deployment framework SGLang has added native decision-making APIs, enabling off-the-shelf LLMs and multimodal models like Qwen3.8-27B to directly perform selection, judgment, and scoring without modifying weights or undergoing retraining. When developers input the current context and several candidate answers, the model directly returns the probability of each option, eliminating the need to first generate a text segment for subsequent parsing. This functionality leverages the "next-token probability" that large language models inherently compute. For example, if candidate answers are A, B, and C, a standard chat model would continue generating text, while SGLang directly reads the model’s probabilities for these options to determine its preferred choice. The model itself remains unchanged; only the method of reading outputs is adjusted. SGLang has demonstrated this capability with a *Pokémon FireRed* demo using Qwen3.8-27B. The model decides whether to attack, switch Pokémon, or heal based on real-time game state, with each decision taking less than 100ms, and successfully cleared the Elite Four and Champion in one run. Since Qwen3.8-27B natively supports image input, this approach can also handle multimodal decision-making. Additionally, SGLang has added a new /v1/systemone endpoint, compatible with the TypeSafe SDK used by Jev. Applications already using this SDK can switch their service address to their own SGLang instance to continue making calls. The new feature is now available in the nightly build and scheduled for official release alongside version v0.5.21.

Needham reaffirms SpaceX’s Buy rating and maintains a $250 price target, with rapid growth in AI computing revenue as the core driver. The firm’s report notes that SpaceX’s announced computing transactions are projected to generate an annualized revenue run rate of $54 billion, strongly supporting management’s goal of hitting $100 billion in total annual recurring revenue (ARR) by the end of 2026. Separately, Starship’s 14th flight successfully reached orbit and deployed 26 Starlink V3 satellites, underscoring SpaceX’s sustained technological edge in rocket reusability and satellite constellation expansion.