←Search results

By relevance · Total 2760 results Clear search

Sunday

  1. 36Kr · AI

    DeepMind has four Gemini agents team up to defeat the Ender Dragon in Minecraft

    Google DeepMind had four Gemini 4 Argon agents team up to play Minecraft from scratch, defeating the Ender Dragon after 27 hours of real time, during which they died 605 times, wrote more than 2,700 programs, and produced a total of 63,300 lines of code. The four agents had no human-written programs to rely on; chopping trees, mining, and crafting equipment all required them to write and test their own code, and they also shared experience with each other and reported the locations where teammates died. The article also reviews work from the 2016 Malmo platform and the 2019 MineRL diamond-mining competition to 2022's OpenAI VPT and 2023's Voyager, among other efforts that use Minecraft to study AI.

    Google DeepMind had four Gemini 4 Argon agents team up to play Minecraft from scratch, defeating the Ender Dragon after 27 hours of real time, during which they died 605 times, wrote more than 2,700 programs, and produced a total of 63,300 lines of code. The four agents had no human-written programs to rely on; chopping trees, mining, and crafting equipment all required them to write and test their own code, and they also shared experience with each other and reported the locations where teammat

    Matched fieldsBody

  2. Shanghai Observer

    Teng Junjie, Gong Tianpeng and Ma Xiaohui Discuss Cultural Going Global: From Peking Opera Films to AI "Horse Racing"

    On October 24, director Teng Junjie, composer Gong Tianpeng and erhu player Ma Xiaohui shared their experiences in the international dissemination of Chinese culture around the essay collection "Wind of the Sea" at the Shanghai Pudong Library. An AI version of the "Horse Racing" MV directed by Teng Junjie was shown on site; the live-action portions were shot in a studio and the rest was generated by AI, greatly reducing costs compared with full live-action shooting. Production of the project began in January this year, and the latest version was just completed two days ago.

    On October 24, director Teng Junjie, composer Gong Tianpeng and erhu player Ma Xiaohui shared their experiences in the international dissemination of Chinese culture around the essay collection "Wind of the Sea" at the Shanghai Pudong Library. An AI version of the "Horse Racing" MV directed by Teng Junjie was shown on site; the live-action portions were shot in a studio and the rest was generated by AI, greatly reducing costs compared with full live-action shooting. Production of the project beg

    Matched fieldsBody

  3. Huxiu

    Sanan Optoelectronics' actual controller Lin Xiucheng arrested on suspicion of embezzlement and misappropriation of funds

    Sanan Optoelectronics announced that its actual controller Lin Xiucheng has been approved for arrest by the procuratorate on suspicion of the crimes of embezzlement and misappropriation of funds. On that day the share price was 10.10 yuan and total market value about 50.4 billion yuan. In 2025 the company's revenue was 17.949 billion yuan, up 11.45% year on year, while net profit attributable to shareholders showed a loss of 353 million yuan, the first annual loss since its backdoor listing in 2008. The 1.312 billion shares held by controlling shareholders Sanan Electronics and Sanan Group have been 100% judicially frozen, accounting for 26.30% of total share capital; publicly disclosed rigid debt is about 12 billion to 15 billion yuan, and local industrial capital in Hubei and Hunan has applied for its restructuring.

    Sanan Optoelectronics announced that its actual controller Lin Xiucheng has been approved for arrest by the procuratorate on suspicion of the crimes of embezzlement and misappropriation of funds. On that day the share price was 10.10 yuan and total market value about 50.4 billion yuan. In 2025 the company's revenue was 17.949 billion yuan, up 11.45% year on year, while net profit attributable to shareholders showed a loss of 353 million yuan, the first annual loss since its backdoor listing in 2

    Matched fieldsBody

  4. Huxiu

    Tested: 6 major AIs each wrote 5 episodes of a restaurant business-war short drama — Claude comes closest to the "heartburn style," while Doubao excels at punchy lines

    After having Doubao, Kimi, DeepSeek, GPT, Qwen and Claude each write 5 episodes of a restaurant business-war short drama, the 30 scripts were highly homogenized: noodle shops appeared 6 times and all six models chose a noodle shop for their first version, "conscience" appeared in every script, and "rules" in about 80%. Claude's dialogue is short and rhythmic, with an apprentice urging a price cut in every version, making it closest to the "heartburn style"; Doubao packs in punchy lines but 3 versions never specify what the shop sells. Across the 30 scripts, "heartburn" does not appear even once.

    After having Doubao, Kimi, DeepSeek, GPT, Qwen and Claude each write 5 episodes of a restaurant business-war short drama, the 30 scripts were highly homogenized: noodle shops appeared 6 times and all six models chose a noodle shop for their first version, "conscience" appeared in every script, and "rules" in about 80%. Claude's dialogue is short and rhythmic, with an apprentice urging a price cut in every version, making it closest to the "heartburn style"; Doubao packs in punchy lines but 3 ver

    Matched fieldsBody

  5. Yicai

    AI Weekly: Meta's Muse surpasses 10 million downloads; CXMT and Micron diverge on technology

    Meta's AI assistant Muse has accumulated 10.9 million downloads since its September launch, of which 10.2 million came from the United States, making it one of the mobile apps to surpass 10 million U.S. downloads within 40 days of launch. DeepSeek open-sourced its inference framework DeepInfer and disclosed that a single cluster serves a throughput of 155 trillion tokens per day, with revenue of $38 per card per hour. Anthropic launched Claude Duet, a personal agent based on Claude Opus 5.5, opening research preview waitlist applications to Claude Pro and Max subscribers in the United States.

    Meta's AI assistant Muse has accumulated 10.9 million downloads since its September launch, of which 10.2 million came from the United States, making it one of the mobile apps to surpass 10 million U.S. downloads within 40 days of launch. DeepSeek open-sourced its inference framework DeepInfer and disclosed that a single cluster serves a throughput of 155 trillion tokens per day, with revenue of $38 per card per hour. Anthropic launched Claude Duet, a personal agent based on Claude Opus 5.5, ope

    Matched fieldsBody

  6. Yicai

    Tech Weekly | Scalpers' markup margins shrink for iPhone Duo; celebrities also struggle to hold AI "face-stealing" livestream sellers accountable

    Apple's first foldable phone, the iPhone Duo, went on sale on October 23, and scalpers' actual markups have shrunk noticeably: the 512GB and 1TB versions carry a markup of about 1,500 yuan, while the 2TB version carries a markup of only a few hundred yuan, even as shipping times on Apple's official website have slipped to January 22. Actress Wu Yue said in a post the same day that someone had used AI to imitate her voice for livestream selling, and demanded that the accounts involved refund the money. Starting from its third cohort, the He Xiangjian Young Scientist Program will add intelligent science as a new funding area.

    Apple's first foldable phone, the iPhone Duo, went on sale on October 23, and scalpers' actual markups have shrunk noticeably: the 512GB and 1TB versions carry a markup of about 1,500 yuan, while the 2TB version carries a markup of only a few hundred yuan, even as shipping times on Apple's official website have slipped to January 22. Actress Wu Yue said in a post the same day that someone had used AI to imitate her voice for livestream selling, and demanded that the accounts involved refund the

    Matched fieldsBody

  7. Wall Street CN

    When Will the AI Trade Stop Falling: The Financing Prisoner's Dilemma Behind Meta's Massive Bond Issuance

    Wall Street CN, drawing on Barclays and Goldman Sachs data, points out that after its third-quarter results, Meta plans to raise hundreds of billions of dollars starting this fall and has narrowed its 2026 capital expenditure forecast to $135 billion to $145 billion, with free cash flow turning negative after dividends. Barclays believes that multiple companies within the AI ecosystem have rushed to launch financing, tightening the capital market's absorption capacity, and that collective bond issuance has instead pushed up issuance costs. The U.S. high-yield bond risk premium has risen to a 17-month high, yields on CCC-rated junk bonds have climbed to 15.85%, and the dispersion of spreads on AI-related credit assets has risen from 24 basis points to 65 basis points over the past year.

    Wall Street CN, drawing on Barclays and Goldman Sachs data, points out that after its third-quarter results, Meta plans to raise hundreds of billions of dollars starting this fall and has narrowed its 2026 capital expenditure forecast to $135 billion to $145 billion, with free cash flow turning negative after dividends. Barclays believes that multiple companies within the AI ecosystem have rushed to launch financing, tightening the capital market's absorption capacity, and that collective bond i

    Matched fieldsBody

  8. TMTPost · AGI

    CATL earned a net profit of 21.2 billion yuan in a single quarter, yet fell nearly 15% in two days: what exactly is the AI rally killing?

    CATL's third-quarter net profit attributable to shareholders was 21.246 billion yuan, up 14.54% year on year, below the lower bound of institutional forecasts. On October 21 its A-shares closed down 12.03%, and over two trading days they fell nearly 15% cumulatively, wiping about 140 billion yuan off its market value. This round of adjustment is weighing on valuations rather than already-realized earnings: the ChiNext Index's PE-TTM fell from about 54 times in June to about 32 times, while the STAR 50 dropped from about 264 times to above 110 times. Over the same period, equity ETFs saw net inflows for eight consecutive trading days, totaling 94.753 billion yuan as of October 14.

    CATL's third-quarter net profit attributable to shareholders was 21.246 billion yuan, up 14.54% year on year, below the lower bound of institutional forecasts. On October 21 its A-shares closed down 12.03%, and over two trading days they fell nearly 15% cumulatively, wiping about 140 billion yuan off its market value. This round of adjustment is weighing on valuations rather than already-realized earnings: the ChiNext Index's PE-TTM fell from about 54 times in June to about 32 times, while the S

    Matched fieldsBody

  9. MarkTechPost

    How to Run Underdog Saluki 27B: A Tool-Calling Guide for 2-bit Qwen3.8-27B

    Underdog Saluki 27B is a 7.9 GB quantized version of Qwen3.8-27B that, according to its developer, retains tool-calling capability at roughly 2 bits per weight and slightly outperforms the full 54 GB model on a self-built 120-task benchmark. The tutorial places Saluki side by side with ISTA-DASLab's GSQ-RCO quantized version in native llama.cpp, compares GGUF headers tensor by tensor and performs byte-level comparisons, then measures memory and speed on a single GPU, and scores both with the official BFCL checker on the same 120 tasks using strict and lenient parsing respectively.

    Underdog Saluki 27B is a 7.9 GB quantized version of Qwen3.8-27B that, according to its developer, retains tool-calling capability at roughly 2 bits per weight and slightly outperforms the full 54 GB model on a self-built 120-task benchmark. The tutorial places Saluki side by side with ISTA-DASLab's GSQ-RCO quantized version in native llama.cpp, compares GGUF headers tensor by tensor and performs byte-level comparisons, then measures memory and speed on a single GPU, and scores both with the off

    Matched fieldsBody

  10. MarkTechPost

    Cognition Open-Sources Devin Vulnerability Fixes Plugin, Covering Three High-Severity Vulnerability Categories in PHP, Node.js, and GitLab

    Cognition has released the open-source, free Devin Vulnerability Fixes plugin, which includes three Agent skills—fix-php-vuln, fix-node-vuln, and fix-gitlab-vuln—under the MIT license, targeting PHP2Shell, the Node.js October security release, and the self-managed GitLab vulnerability CVE-2026-110264 with a CVSS score of 10.0. The skills execute in five steps: read-only inventory, approval, target version verification, before-and-after testing, and one PR per repository. They can run in agents that read SKILL.md, such as Devin, Claude Code, Codex, Cursor, and GitHub Copilot, and require git, shell, Docker, and image registry access; npx installation requires Node.js 22.20 or later. The original article does not publish a fix-rate benchmark; hosted runtimes, all-in-one machines, and Electron applications are only flagged and not patched.

    Cognition has released the open-source, free Devin Vulnerability Fixes plugin, which includes three Agent skills—fix-php-vuln, fix-node-vuln, and fix-gitlab-vuln—under the MIT license, targeting PHP2Shell, the Node.js October security release, and the self-managed GitLab vulnerability CVE-2026-110264 with a CVSS score of 10.0. The skills execute in five steps: read-only inventory, approval, target version verification, before-and-after testing, and one PR per repository. They can run in agents t

    Matched fieldsBody

  11. Phoenix Technology

    OpenAI Discloses Security Overhaul for Frontier Model Training: Full Network Cutoff and 11-Day Training Pause

    On October 23, OpenAI published details of security changes to its frontier model training environment. Frontier model training is no longer allowed to access the real internet; instead, models practice browsing and search in an offline web environment. In red-team experiments, an Agent, under conditions prohibiting network communication and with no shared writable files, used side channels such as CPU caches to exchange information; in another test, an internal research model exploited a DNS filtering gap to contact an external chatbot. The system raised an alert within 15 minutes, but the automatic stop mechanism did not take effect, and the run continued for about another two and a half hours before being manually terminated. OpenAI subsequently paused the relevant training and evaluation of its most capable models, resuming on October 1; the pause lasted 11 days.

    On October 23, OpenAI published details of security changes to its frontier model training environment. Frontier model training is no longer allowed to access the real internet; instead, models practice browsing and search in an offline web environment. In red-team experiments, an Agent, under conditions prohibiting network communication and with no shared writable files, used side channels such as CPU caches to exchange information; in another test, an internal research model exploited a DNS fi

    Matched fieldsBody

  12. Phoenix Technology

    AI short dramas may not be "compute slop": they may be creating new demand for compute

    The widely circulated claim that "AI short dramas and video generation account for 55% of domestic token consumption" is inaccurate; underlying logs cannot distinguish whether generated content belongs to short dramas or advertisements. Data from the China Netcasting Services Association shows that in the first half of 2026, platforms launched approximately 367,000 micro short dramas, of which about 270,000 were AI micro short dramas, accounting for over 74%; Douyin launched 221,900 AI dramas in the first half of the year, with only 1,055 surpassing 100 million views. The high-frequency video generation and repeated rerolling brought by AI short dramas are creating new inference demand, rather than simply crowding out existing compute for programming and scientific research.

    The widely circulated claim that "AI short dramas and video generation account for 55% of domestic token consumption" is inaccurate; underlying logs cannot distinguish whether generated content belongs to short dramas or advertisements. Data from the China Netcasting Services Association shows that in the first half of 2026, platforms launched approximately 367,000 micro short dramas, of which about 270,000 were AI micro short dramas, accounting for over 74%; Douyin launched 221,900 AI dramas in

    Matched fieldsBody

  13. Sina Tech

    OpenAI's price cuts pay off in winning orders: multiple Anthropic customers shift coding business to OpenAI

    OpenAI attracted Anthropic customers through large discounts and free credits. In Rippling's coding tasks, the share of OpenAI models has risen from 16% to 52%, while Anthropic models fell from 57% to 18%. Replit moved most of its AI budget from Anthropic to OpenAI; OpenAI offered a discount of over 30%, and the annual contract size exceeds $100 million. After Anthropic released Claude Opus 5.5 at the end of last month, it reclaimed the top spot in usage share on OpenRouter, and over the past week its popularity was about twice that of OpenAI models.

    OpenAI attracted Anthropic customers through large discounts and free credits. In Rippling's coding tasks, the share of OpenAI models has risen from 16% to 52%, while Anthropic models fell from 57% to 18%. Replit moved most of its AI budget from Anthropic to OpenAI; OpenAI offered a discount of over 30%, and the annual contract size exceeds $100 million. After Anthropic released Claude Opus 5.5 at the end of last month, it reclaimed the top spot in usage share on OpenRouter, and over the past we

    Matched fieldsBody

  14. Huxiu

    OpenAI Publishes 722 Mathematics Papers at Once, Mathematicians Worry Research Will Be Flattened by AI

    OpenAI published 722 mathematics papers at once, covering number theory, geometry, and theoretical computer science, claiming progress on problems such as the Riemann Hypothesis, Catalan's constant, and the Kakeya conjecture, without peer review. Fields Medalist Hugo Duminil-Copin said 16 problems related to his research had been flattened by AI, and some doctoral students face their topics being covered and their degrees affected. The article also mentions an Anthropic researcher claiming to have disproved the Jacobian conjecture, and Shing-Tung Yau leading the establishment of a mathematics and artificial intelligence laboratory alliance.

    OpenAI published 722 mathematics papers at once, covering number theory, geometry, and theoretical computer science, claiming progress on problems such as the Riemann Hypothesis, Catalan's constant, and the Kakeya conjecture, without peer review. Fields Medalist Hugo Duminil-Copin said 16 problems related to his research had been flattened by AI, and some doctoral students face their topics being covered and their degrees affected. The article also mentions an Anthropic researcher claiming to ha

    Matched fieldsBody

  15. East Money

    Brokerages say the market can recover, focus on investment opportunities in physical AI, solid-state batteries, innovative drugs, and more

    Huafu Securities believes that the valuation of A-share core assets has fully priced in pessimistic expectations, with limited downside, and recommends avoiding short-term disturbances and focusing on targets whose performance and valuation are matched. Southwest Securities says solid-state batteries are expected to enter the pilot testing and small-batch mass production window in 2026–2027, with investment opportunities to be realized first in equipment and materials. Guosheng Securities notes that at the 2026 ESMO Annual Meeting, a total of 28 studies from China were selected for LBA, of which 4 entered the plenary session.

    Huafu Securities believes that the valuation of A-share core assets has fully priced in pessimistic expectations, with limited downside, and recommends avoiding short-term disturbances and focusing on targets whose performance and valuation are matched. Southwest Securities says solid-state batteries are expected to enter the pilot testing and small-batch mass production window in 2026–2027, with investment opportunities to be realized first in equipment and materials. Guosheng Securities notes

    Matched fieldsBody

  16. East Money

    Wall Street Earnings Season Peaks: Can AI Trades Withstand Fed Rate Hike Pressure

    Next week, about one-third of S&P 500 companies, including Alphabet, Microsoft, and Amazon, will report earnings, while the Federal Reserve will hold its policy meeting. Markets expect only about a 10% probability of a rate hike this month and about 80% in December. S&P 500 third-quarter earnings are expected to grow more than 32%, but investors are pulling out of AI trades on concerns about AI capital expenditures, with the communication services sector down 6.2% over the past week. Bank of America recommends maintaining a tactical risk-off stance until mid-November.

    Next week, about one-third of S&P 500 companies, including Alphabet, Microsoft, and Amazon, will report earnings, while the Federal Reserve will hold its policy meeting. Markets expect only about a 10% probability of a rate hike this month and about 80% in December. S&P 500 third-quarter earnings are expected to grow more than 32%, but investors are pulling out of AI trades on concerns about AI capital expenditures, with the communication services sector down 6.2% over the past week. Bank of Ame

    Matched fieldsBody

  17. Yicai

    Wall Street Earnings Season Peaks: Can the AI Trade Withstand Fed Pressure?

    U.S. stocks are set for a wave of earnings reports and a Federal Reserve rate meeting, with markets focused on the costs and return risks of AI investment. Third-quarter earnings for S&P 500 companies are expected to grow more than 32%, and capital expenditure forecasts from Alphabet, Microsoft and Amazon are in focus. Markets expect about a 10% probability of a Fed rate hike this month and about an 80% probability for December.

    Matched fieldsBody

  18. Phoenix Technology

    Apple's foldable screen cracks; large model prices rebound; Xiaoyunque merges into Jimeng; Google illegally cuts down trees

    Users of Apple's iPhone Duo foldable have reported that the screen cracks right out of the box, with the failures mostly concentrated directly above the hinge and manifesting as a bright pixel line appearing on one side. A CAICT report shows that domestic token prices have doubled compared with the second half of last year, with the normalized average price rising from 6.8 yuan to 13.46 yuan, but average daily consumption is about 500 trillion, nearly 4 times that of December 2025. ByteDance has merged its AI short-drama creation platform Xiaoyunque into Jimeng, and the product retains a standalone entry point.

    Users of Apple's iPhone Duo foldable have reported that the screen cracks right out of the box, with the failures mostly concentrated directly above the hinge and manifesting as a bright pixel line appearing on one side. A CAICT report shows that domestic token prices have doubled compared with the second half of last year, with the normalized average price rising from 6.8 yuan to 13.46 yuan, but average daily consumption is about 500 trillion, nearly 4 times that of December 2025. ByteDance has

    Matched fieldsBody

Saturday

  1. Analytics India Magazine

    DrizzVision's AI App Testing Ambitions and Reality

    DrizzVision has launched Fathom, an AI agent that uses vision models to understand app interfaces and execute mobile testing according to natural-language instructions. The company says that on its internal benchmarks, test flakiness dropped from 15%–50% to about 3%. When testing 6 frontier model configurations across 20 tasks on Android, the standard image suite achieved a 72.3% pass rate, while the harder suite on the rooted image achieved only 28.7%. GPT 5.6 Luna fell from 64% to 4%, while Gemini 3.5 Flash rose from 54% to 34%.

    DrizzVision has launched Fathom, an AI agent that uses vision models to understand app interfaces and execute mobile testing according to natural-language instructions. The company says that on its internal benchmarks, test flakiness dropped from 15%–50% to about 3%. When testing 6 frontier model configurations across 20 tasks on Android, the standard image suite achieved a 72.3% pass rate, while the harder suite on the rooted image achieved only 28.7%. GPT 5.6 Luna fell from 64% to 4%, while Ge

    Matched fieldsBody