From 19 items, 8 important content pieces were selected
- China’s first mandatory L3/L4 autonomous driving standard submitted for approval, to take effect in 2027 ⭐️ 9.0/10
- 3D optical fiber micro-tweezer achieves 100,000x force ⭐️ 9.0/10
- SMPTE Opens Standards Library for Free Access ⭐️ 8.0/10
- Cloudflare Launches Temporary Worker Accounts for AI Agents ⭐️ 8.0/10
- Anthropic to Meet White House Over AI Model Shutdown ⭐️ 8.0/10
- IETF Proposes HTTP QUERY Method for Safe Queries with Body ⭐️ 8.0/10
- LM Studio and Apple Run 1T-Parameter Model on Four Mac Studios ⭐️ 8.0/10
- Tencent to embed AI agent in WeChat, compliance approval this month ⭐️ 8.0/10
China’s first mandatory L3/L4 autonomous driving standard submitted for approval, to take effect in 2027 ⭐️ 9.0/10
China’s Ministry of Industry and Information Technology (MIIT) has completed the draft of the mandatory national standard ‘Safety Requirements for Intelligent and Connected Vehicle Autonomous Driving Systems’ and submitted it for approval. It is recommended to take effect on July 1, 2027, introducing a safety case mechanism and hard safety requirements for L3 and L4 autonomous driving. This is the first mandatory national standard for L3/L4 autonomous driving in China, shifting regulation from conceptual easing to hard safety constraints. It will end vague marketing claims by automakers and shift competition toward safety capabilities, potentially raising short-term costs but promoting long-term innovation. The standard introduces a Safety Case mechanism requiring companies to systematically demonstrate safety through ‘claim-argument-evidence’ reasoning. It also imposes specific requirements for L3 human-machine handover and L4 autonomous risk handling.
telegram · zaihuapd · Jun 20, 03:31
Background: Autonomous driving levels are defined by SAE International: L3 (conditional automation) requires the driver to be ready to take over, while L4 (high automation) can handle all driving tasks in specific conditions without human intervention. A safety case is a structured argument supported by evidence that a system is acceptably safe for a given application. China has been developing autonomous driving regulations to catch up with technological advancements and global standards.
References
Tags: #autonomous driving, #regulation, #China, #L3/L4, #safety standard
3D optical fiber micro-tweezer achieves 100,000x force ⭐️ 9.0/10
Researchers from Anhui University and University of Science and Technology of China have developed a novel 3D optical fiber micro-tweezer using femtosecond laser composite manufacturing, achieving over 100,000 times the output force of traditional optical tweezers. The device integrates light transmission, photothermal conversion, material response, and microstructural mechanical output into a single optical fiber, enabling high-precision, low-damage, and programmable 3D manipulation of microscale objects. This breakthrough, published in Nature, overcomes key limitations of traditional optical tweezers such as weak force and inability to manipulate opaque objects. It provides a powerful new tool for single-cell biology and minimally invasive medicine, with potential to revolutionize micromanipulation in biomedical research and clinical applications. The micro-tweezer achieves continuous, precise force control by adjusting input light power, and can perform accurate sampling in spaces as small as 100 micrometers. Unlike traditional optical tweezers that rely on gradient forces for transparent particles, this device utilizes photothermal effects and mechanical deformation of the fiber tip to generate forces, enabling manipulation of opaque objects.
telegram · zaihuapd · Jun 20, 15:19
Background: Traditional optical tweezers use highly focused laser beams to trap and manipulate microscopic particles, but they typically generate very small forces (piconewtons) and can only handle transparent or dielectric objects. The new 3D optical fiber micro-tweezer overcomes these limitations by integrating multiple functionalities into a single fiber, leveraging femtosecond laser fabrication to create complex microstructures on the fiber tip. This approach allows for much larger forces and broader applicability, including manipulation of biological cells and opaque particles.
References
Tags: #optical tweezers, #fiber optics, #biophysics, #nanophotonics, #laser micro-fabrication
SMPTE Opens Standards Library for Free Access ⭐️ 8.0/10
The Society of Motion Picture and Television Engineers (SMPTE) has announced that its entire library of media technology standards is now freely accessible to the public, removing previous paywalls. This move removes a major barrier to innovation in media production and distribution, enabling broader adoption and collaboration on open standards across the industry. The free access includes standards such as SMPTE timecode and VC-1, and is part of SMPTE’s modernization efforts including GitHub workflows and structured HTML authoring.
hackernews · zdw · Jun 20, 17:01 · Discussion
Background: SMPTE is a professional society that develops technical standards for the motion picture and television industries, including the widely-used SMPTE timecode. Previously, access to these standards required payment, which limited adoption and innovation.
References
Discussion: Comments largely welcomed the move, with lambdaone praising it as aiding the shift to open standards like the IETF, while geerlingguy questioned why any standards body wouldn’t do this by default. ksec noted the modernization initiatives like GitHub adoption.
Tags: #standards, #media technology, #open access, #SMPTE, #industry news
Cloudflare Launches Temporary Worker Accounts for AI Agents ⭐️ 8.0/10
Cloudflare has introduced temporary Worker accounts that allow AI agents and developers to deploy Workers via ‘wrangler deploy –temporary’, creating live deployments that last 60 minutes. These temporary accounts can be claimed to become permanent, or they expire automatically. This feature removes the human authentication barrier for AI agents, enabling fully automated ephemeral deployments for testing, PR previews, and AI-driven workflows. It also provides free 60-minute scratch deployments for any user, which is valuable for experimentation. The temporary deployment is live for 60 minutes, after which it expires unless claimed via a permanent Cloudflare account. Cloudflare applies rate limits and abuse prevention checks to prevent misuse, such as limiting how quickly temporary preview accounts can be created.
hackernews · farhadhf · Jun 20, 11:19 · Discussion
Background: Cloudflare Workers is a serverless platform that runs JavaScript and other languages on edge servers. Previously, deploying a Worker required human interaction for authentication (OAuth, MFA, API tokens). Temporary Accounts automate this process, enabling AI agents to deploy without human involvement, aligning with the growing trend of ephemeral serverless deployments.
References
Discussion: Community members expressed excitement about ephemeral deployments for PR previews and code review, but also raised concerns about the lack of hard billing caps and potential abuse of temporary infrastructure. Some users wished for better copywriting in the announcement.
Tags: #Cloudflare Workers, #ephemeral deployments, #AI agents, #serverless
Anthropic to Meet White House Over AI Model Shutdown ⭐️ 8.0/10
Anthropic senior technical staff are scheduled to meet with White House officials next week to resolve a dispute that led to the global shutdown of their most advanced AI models, Fable 5 and Mythos 5. The shutdown followed a Trump administration order blocking foreign access to these models. This meeting represents a direct government intervention in AI deployment, underscoring tensions between national security and AI accessibility. The outcome could set a precedent for how advanced AI models are regulated globally, affecting industry practices and international norms. The Trump administration ordered Anthropic to prevent foreign users from accessing Fable 5 and Mythos 5, leading to a global ban. Mythos 5 was previously delayed due to its potent hacking capabilities, particularly its ability to find and exploit zero-day vulnerabilities.
telegram · zaihuapd · Jun 20, 02:45
Background: Anthropic recently launched Claude Fable 5 as a public, safe version of the more powerful Mythos 5 model, which had raised cybersecurity concerns due to its ability to autonomously discover and exploit vulnerabilities. The White House order likely reflects fears that foreign adversaries could misuse such capabilities for offensive cyber operations. This situation highlights the growing friction between AI safety measures and government security mandates.
References
Tags: #AI regulation, #Anthropic, #government policy, #AI safety, #model shutdown
IETF Proposes HTTP QUERY Method for Safe Queries with Body ⭐️ 8.0/10
The IETF HTTP Working Group is developing a draft to introduce a new HTTP QUERY method, allowing query parameters to be sent in the request body while maintaining safe and idempotent semantics similar to GET. The draft also defines a new Accept-Query response header for servers to advertise supported query formats. This proposal addresses the long-standing limitation of URI length by enabling complex queries to be sent in the request body, which is crucial for modern web applications with large query payloads. It also improves cacheability and reliability compared to POST-based workarounds, potentially impacting how APIs and web services handle query operations. The QUERY method is safe and idempotent, meaning it does not change server state and repeated requests produce the same result, enabling automatic retries and caching. The current draft (version -14) expires on December 22, 2026, and is still under active discussion within the working group.
telegram · zaihuapd · Jun 20, 06:28
Background: HTTP GET requests pass parameters via the URI query string, which is limited in length (usually 2-8 KB depending on the client and server). POST requests can carry a body but are not safe or idempotent by default, making them unsuitable for caching and automatic retries. The proposed QUERY method fills this gap by combining the safety/idempotence of GET with the ability to send a request body, akin to how some APIs already use POST for queries but without the standard guarantees.
References
Tags: #HTTP, #protocol, #IETF, #web development, #query method
LM Studio and Apple Run 1T-Parameter Model on Four Mac Studios ⭐️ 8.0/10
At WWDC, LM Studio and Apple demonstrated running the 1 trillion parameter Kimi K2.6 model on a cluster of four Mac Studios using a preview version of LM Studio, and also showcased secure remote access via LM Link from a MacBook and iPhone. This demonstration highlights the feasibility of running massive models on consumer hardware via distributed inference, potentially democratizing access to large language models without expensive dedicated servers. The model used is Kimi K2.6, an open-source multimodal agentic model by Moonshot AI with strong coding capabilities. LM Link enables remote access to local models across devices, making it possible to use a laptop or phone to interact with a model running on a powerful cluster.
telegram · zaihuapd · Jun 20, 07:02
Background: LM Studio is a beginner-friendly desktop application for running local AI models on personal computers without requiring command-line skills. Apple’s Mac Studio features unified memory which allows large models to be loaded across multiple machines. Distributed inference techniques split a model across multiple devices to handle larger models than a single device can support.
References
Tags: #LM Studio, #Apple, #Mac Studio, #large language model, #distributed inference
Tencent to embed AI agent in WeChat, compliance approval this month ⭐️ 8.0/10
Tencent is testing a prototype of an AI agent embedded in WeChat, with plans to begin compliance approval as early as this month, followed by small-scale external testing and phased rollout. Users will be able to swipe right to access the agent and issue voice or text commands to complete tasks automatically via mini-programs. Given WeChat’s massive user base (over 1.3 billion monthly active users), this integration could make AI agents mainstream in China, enabling hands-free task completion within the country’s most popular super-app. It also intensifies competition with Alibaba and ByteDance, which have already launched agent features in their respective apps (Tongyi and Doubao). Tencent faces a computing power bottleneck due to limited reserves of NVIDIA chips and tight domestic semiconductor supply, making large-scale rollout costly and near-term profitability uncertain. The agent will reportedly trigger mini-programs to fulfill user requests, with an example being ordering coffee based on taste and price preferences.
telegram · zaihuapd · Jun 20, 09:23
Background: An AI agent is an autonomous system that uses large language models to understand user requests and perform multi-step actions, such as booking tickets or ordering food, by interacting with other software. WeChat is China’s dominant social and payments platform, making it a strategic gateway for AI services. Alibaba and ByteDance have already integrated agent features into Tongyi and Doubao, which have seen rapid monthly active user growth, pressuring Tencent to accelerate its plans.
References
Tags: #WeChat, #AI agent, #Tencent, #AI assistants, #mobile platforms