Kling AIFirst-party
AI score22Kling 4.0 Generates New Emil Heliot Campaign Film
#Industry news#Video#Multimodal
NAll news
All news
- TMTPost · AGIResearchAI score70
DeepSeek Harness Team Releases Super-agent Design Exploration and Open-Sources Prototype
On the evening of October 31, the DeepSeek Harness team published a technical article titled "Super-agent Design Exploration: Toward a More Autonomous Agent System" on its official WeChat account, and the accompanying prototype dsh-superagent-prototype was open-sourced the same day on GitHub under the MIT license. The article proposes using a continuously running router to uniformly receive messages from group chats, GitHub, email, and other sources and dispatch tasks, with workers that each maintain independent contexts executing tasks in parallel, and introduces mechanisms such as event-driven operation, separation of storage and compute, watchdog, and two-level message filtering. Preliminary experiments presented in the article show a median task completion time of 12.4 minutes and a 95th percentile of 15.7 minutes, but the authors state that these are preliminary experiments based on synthetic workloads and a simplified environment, and that the GitHub prototype is a research prototype rather than a formal product.
#Product updates#Agents#Open-source ecosystem
- TMTPost · AGIAI score40
In the inference era, the computing power industry is changing its accounting books: Token cost and engineered delivery become the new yardsticks
Inference computing demand has already surpassed training. In 2026, inference computing is projected to account for 70.5% of total AI computing load, and global inference spending will reach $23.3 billion, surpassing training spending of $19 billion for the first time. Customer procurement metrics are shifting from chip models and peak FLOPS to tokens per second, inference latency per unit cost, and operations and maintenance costs. The cost per million tokens on DeepSeek's official website has already dropped from about 130 yuan at the beginning of the year to under 3 yuan. The focus of industry competition has accordingly shifted to engineered delivery and operations and maintenance, with heterogeneous computing and ASIC inference chips becoming the new battleground.
#English text is being prepared.#Reasoning#English text is being prepared.
- Hacker News trendingAI score45
Can a modern chess engine running on a 1996 Pentium beat 1996 Kasparov?
Nicholas Carlini ported Stockfish to 1996 hardware—a 200 MHz Intel Pentium Pro with 64MB of memory—and estimates its classical time-control Elo at 3300 ± 87, far above Kasparov's roughly 2800. The experiment shows that of the +1335 Elo improvement of the modern engine over 1996 Crafty, about +720 comes from algorithmic and programming improvements rather than hardware. The engine played 100 games against 1996 Crafty on modern hardware, scoring 78 wins, 17 draws, and 5 losses.
#Papers/Research#Open-source ecosystem
- Hacker News TrendingAI score70
Revolut Migrates About 2,000 Engineers to Self-Hosted GLM for Coding, Claude Code and Codex Become Exceptions
According to Guillermo Flor, relaying statements from Revolut AI operations lead David Abitbol, Revolut has migrated the coding work of about 2,000 engineers to self-hosted GLM and placed Claude Code and Codex on an exception list. Its configuration is 56 Nvidia B200 GPUs, with a blended cost of about 30 cents per million tokens, roughly one-tenth of what it previously paid Anthropic and OpenAI, and engineering delivery cycles improved by 70%. The author believes that once tokens were no longer rate-limited, engineers applied AI to all tasks, and that frontier models are becoming the exception while open weights become the default choice.
#Industry news#English text is being prepared.#Open-source ecosystem
- East MoneyAI score31
Yan Xiang of Founder Securities: Watch Global AI Capital Expenditure; Bullish on Two Main Lines for A-Shares in Q4—Hardcore Tech and Undervalued Dividend Stocks
Yan Xiang, Chief Economist at Founder Securities, believes that the foundation for the medium- and long-term improvement of A-shares in Q4 remains unchanged, and that investors can focus on two main lines: the hardcore tech sector with strong certainty of earnings realization, and undervalued, high-dividend dividend assets. He noted that the Fed's 25-basis-point rate hike in September to 3.75%-4%, combined with doubts about the sustainability of AI capital expenditure, caused a significant pullback in global tech stocks in Q3; the median forward P/E of the seven U.S. AI hardware companies peaked in October 2025, capital expenditure growth of about 90% in the first half was the highest of this cycle, and the market expects a slowdown in 2027.
#Industry news#English text is being prepared.
- Smart City Industry AnalysisAI score48
NDRC "15th Five-Year Plan" Press Conference: Key Work Points on Emerging Pillar Industries, AI+, and Data Elements
At the first press conference of the "Striving for the 15th Five-Year Plan," the National Development and Reform Commission (NDRC) stated that by the end of the 15th Five-Year Plan period, the output value of emerging pillar industries is expected to double to more than 10 trillion yuan, and made clear that the "AI+" initiative will be deepened and expanded. On the AI front, more than 8,000 primary-level medical institutions in China have already connected to cloud-based AI imaging services, and urban unmanned patrol vehicles can identify more than 40 types of city appearance issues. The National Data Administration said it will improve data property rights registration, build a nationally integrated data market, and plan a new round of the three-year "Data Elements ×" initiative.
#English text is being prepared.#Industry news
- English text is being prepared.AI score28
JD Double 11 Report: Waveguide Smart Glasses Sales Up More Than 3x Year-on-Year, RayNeo iO Ranks First in Single-Item Sales
JD's Double 11 opening report shows that from October 12 to November 1, sales of waveguide smart glasses grew more than 3x year-on-year. RayNeo simultaneously topped JD's two major brand rankings for smart glasses and XR devices, and the new RayNeo iO ranked first among waveguide glasses in single-item sales. RayNeo GT Max ranked first among AR glasses in single-item sales; its equivalent 6-meter 267-inch giant screen and native 3DoF spatial hovering continue the momentum from 618. During the event, the final price of RayNeo iO after stacking discounts and national subsidies was 1,911 yuan.
#Industry news#English text is being prepared.
- Semiconductor Industry ObserverAI score46
TSMC Capacity Tight, Samsung Wins US Cloud Provider's 2nm AI Chip Order via Faraday Through DSP Channel
Samsung Electronics has secured a 2nm GAA cloud AI chip ASIC project from a US cloud provider through its Taiwan-based DSP partner Faraday, with the contract value reaching tens of millions of dollars and first revenue expected to be recognized in the first half of 2027. The customer initially planned 4nm, but after weighing performance and power consumption shifted to Samsung's 2nm GAA process; the chip will use 3D-stacked custom DRAM and hybrid bonding packaging. This is Faraday's largest single project since its founding, and also Samsung's latest foothold in expanding HPC customers through the DSP channel, following direct signings with Tesla and Broadcom.
#Industry news#English text is being prepared.
- Phoenix TechnologyAI score40
Intel and Tencent Cloud Release New-Generation Cloud Instances Based on Xeon 6+
At the 2026 Tencent Global Digital Ecosystem Summit, Intel and Tencent Cloud released new-generation cloud instances based on Xeon 6+ processors, delivering an overall performance improvement of more than 30% over the previous generation, with up to 288 efficiency cores and a single processor capable of supporting the stable deployment of more than 1,000 agents. The two parties also demonstrated practices including IAA in-memory compression acceleration (accelerating average decompression time by 16x) and concurrent creation of 350 sandboxes on a single CubeSandbox machine (an improvement of about 46% over other mainstream platforms), and previewed the next-generation Xeon processor, Diamond Rapids.
#Product updates#English text is being prepared.#Agents
- Phoenix TechnologyAI score36
Standard Chartered CEO Winters on AI's Impact on Jobs: Fear Is Real, Training Is the Cure
Standard Chartered CEO Winters, responding to AI's impact on jobs at the Hong Kong Monetary Authority's Global Financial Leaders' Investment Summit, said employees' sense of panic is real, but reskilling and training can ease concerns. Standard Chartered currently has about 100,000 employees, roughly the same as 11 years ago; during that period, automation cut about 25,000 positions while creating new ones. In May, the bank announced plans to cut more than 15% of back-office roles by 2030, about 8,000 support positions. Barclays also plans to cut 5,000 jobs to reduce costs with AI.
#Industry news#English text is being prepared.
- cnBetaAI score56
SUSE announces layoffs of 10%, affecting about 190 employees, with AI and automation listed as the main reasons
SUSE announced a global layoff of 10%, which based on approximately 1,900 employees could affect about 190 people, and plans to begin notifying affected employees starting November 4. CEO van Leeuwen said in an internal memo that AI capabilities are developing at an exponential rate, modern tools have changed internal working methods, and demand for some positions has dropped significantly. Previously, SUSE launched a voluntary departure program in September 2026 but participation fell short of expectations, and the company had also just added agent features to its Linux management tools on October 27, 2026.
#Industry news#Open-source ecosystem
- cnBetaProduct releaseAI score67
CentOS Co-Founder Used GPT-5.6 Sol to Drive Codex to Write the Rust Kernel Ginux in 11 Days
CentOS co-founder Gregory Kurtzer has announced a new kernel project, Ginux, in which memory management, process scheduling, file systems, and networking are all written from scratch in Rust, and existing programs can run without modification by emulating the Linux ABI. The project is driven by a continuously running OpenAI Codex task using GPT-5.6 Sol, which generated approximately 139,000 lines of code and test content and completed more than 1,000 commits within 11 days. Using the x86-64 interface of Linux 6.12 as a reference, of 375 system calls only 82 are fully implemented, 245 are partially available, and 48 are unavailable; it currently runs only in QEMU virtual machines, does not yet support booting on physical hardware, GPUs, or NVMe, and has not undergone an independent security audit.
#English text is being prepared.#Agents#English text is being prepared.
- Sina TechAI score35
DeepSeek down? Official response: partial outage fixed, monitoring repair results
Since 16:58 on November 3, partial outages occurred on the DeepSeek website and API, with affected services including chat, the DeepSeek V4 Pro API, and the DeepSeek V4.1 Flash API. Officials said repairs have been implemented for the outage and the results are being monitored, but the above API services and chat service still show degraded performance.
#Industry news#DeepSeek
- Sina TechAI score40
Li Jiaqi's livestream sells nearly 100,000 Qwen Office memberships in half an hour, corresponding to about 2 trillion Tokens
Li Jiaqi and Chen Yusen, Vice President of Alibaba Group and CEO of Qwen Office, sold Qwen Office personal membership monthly cards in the livestream. 50,000 were sold within 10 minutes of going on sale, and nearly 100,000 cumulatively in about half an hour, corresponding to a quota of about 2 trillion Tokens. The monthly card was priced at 1 cent, and the livestream also showcased products such as AI computers, AI voice-recording cards, and desktop robots.
#Industry news#Agents
- Wired AIAI score39
WIRED Exclusive Interview with New York City Mayor Zohran Mamdani: On AI and Data Centers
In a WIRED exclusive interview, New York City Mayor Zohran Mamdani discussed AI and data centers, coining the phrase "Our Risk, Their Reward." He gave his first interview before being elected in October 2025; now, 10 months into his term, he was interviewed at New York City Hall, with a voter approval rating of 69%.
#Industry news#English text is being prepared.#OpenAI
- heise onlineAI score33
UCLA and UC Davis develop a 185mg all-polymer soft flying robot that can keep flying after being swatted
The UCLA and UC Davis team has developed a 185mg all-polymer micro flying robot that can continue flying without damage after being swatted with a flyswatter or withstanding pressure 11,000 times its own weight. It relies on relaxor ferroelectric polymer artificial muscles to drive flexible conical wings, achieves lift three times its own weight, and can take off autonomously at 0.4 m/s. The results were published in Science Robotics. It currently still relies on external power, and the team plans to add sensors and an onboard power supply.
#Papers/Research#Embodied AI
- heise onlineAI score25
Vitest Browser Mode Replaces jsdom: Frontend Testing with Real Pixels
In an interview, Alexander Opalic explains how he uses Vitest Browser Mode to render Datepicker, tables, and Modal in isolation within an iFrame, testing from the user's perspective rather than through method calls, which can catch issues like gray text on a white background that simulated environments cannot detect. The trade-off is that each click takes 200 milliseconds instead of 20 milliseconds. He also discusses doing screenshot testing for his own UI library, and why it is more worthwhile to establish a clear testing strategy as more and more code is generated by AI.
#Tutorials / practice#English text is being prepared.
- heise onlineAI score41
Cisco has a 27-year-old employee refactor legacy code with AI, aiming to rewrite or refactor 70% of its products by the end of 2027
Cisco President Jeetu Patel assigned 27-year-old Padmanabhan Krishnamurthy to refactor legacy code with AI, giving him unlimited tokens and a "Go nuts" mandate. The project lasts 24 months and aims to rewrite or refactor 70% of products by the end of 2027, rather than rewriting all 98% of the code as previously reported by Fortune. The team consists of one-third senior experts, one-third external market experts, and one-third AI-savvy developers and former founders.
#Industry news#English text is being prepared.#English text is being prepared.
- heise onlineAI score22
Commentary: AI Is a Fever Dream | heise online
A heise commentary argues that although AI has achieved milestones in areas such as protein folding and tumor detection, the other side is more glaring: a Kapwing analysis last year showed that about one-fifth of the videos YouTube shows to new users are AI-generated. The author states that OpenAI lost 3.85 billion US dollars in 2025 and Anthropic lost 4.2 billion US dollars, and criticizes the AI giants for relying on manufacturing hype to maintain investor confidence.
#English text is being prepared.#Industry news#OpenAI