NAll news

All news

Sunday264 items

  1. Phoenix TechnologyResearch
    AI score69

    DeepSeek proposes Super-agent architecture: a two-tier design of Router and Worker

    The official DeepSeek Harness team released "Super-agent Design Exploration," proposing a two-tier architecture in which a Router layer uniformly handles human interaction and task dispatch while multiple Workers execute concurrently in isolated sessions, and summarizing four characteristics: continuous, efficient, proactive, and persistent. Experiments show that Super-agent's task completion time p50 and p95 were 12.4 minutes and 15.7 minutes respectively, while the single-Session approach under neutral prompts had a p95 of 29.5 minutes; the team noted that the experiments were based on synthetic workloads and a simplified environment, mainly evaluating scheduling and collaboration capabilities. The team also provided a prototype implementation named dsh-superagent-prototype, and made clear that it is intended to validate the technical approach and is not an official DeepSeek product.

    Same story · 23 reports

    #Papers/Research#Agents#DeepSeek

  2. Phoenix Technology
    AI score68

    Blackstone Reverses Stance, Says Anthropic May Struggle to Obtain Investment-Grade Credit Rating After IPO

    According to a report by the UK's Financial Times, Blackstone Group has told investors that Anthropic is more likely to receive a speculative-grade rating in the BB range after its IPO, rather than the investment-grade rating previously expected. This reversal emerged in Anthropic's $63 billion chip financing deal; a speculative-grade rating would restrict investment by institutions such as insurance companies and pension funds. The prospectus shows that Anthropic plans to invest more than $800 billion in computing infrastructure by 2045, and currently holds more than $112 billion in cash and cash equivalents.

    Same story · 4 reports

    #Industry news#Anthropic

  3. Phoenix Technology
    AI score68

    Microsoft, Google, Oracle and Two Other Tech Giants Commit Over $3 Trillion to AI Infrastructure

    According to Bloomberg, five major tech giants—Alphabet, Meta, Microsoft, Amazon and Oracle—have committed to investing more than $3 trillion in data centers and other areas over the coming years. Among them, Alphabet's total contract commitments exceed $1 trillion, nearly eight times the level a year earlier; Meta's total future spending exceeds $800 billion, nearly half of which is for data center leases that have not yet taken effect and are payable over terms of up to 30 years. The five companies' combined future lease payment commitments total $1.3 trillion, and other purchase commitments total $1.9 trillion. Oracle has already seen negative free cash flow due to its investment in AI data centers.

    Same story · 3 reports

    #Industry news#English text is being prepared.#OpenAI

  5. Ado Kukic
    AI score20

    Anthropic will support Claude Duet coordinating CC sessions

    In response to user feedback, the Claude Duet team said that users will soon be able to use Claude Duet to coordinate CC sessions, projects, and local threads. Ado relayed this news, saying that he is often asked about this question.

    Quoted postkaran ahuja · @ahuja_karan

    @embw_l0x @AnthropicAI 👋 from Claude Duet team! Appreciate the feedback and hear your frustration. You’ll soon be able to coordinate with CC sessions, projects, and local threads. Stay tuned.

    #Product updates#Agents#English text is being prepared.

  6. indigo
    AI score50

    Grok Bot Team Management Guide: Fewer but Better, Save Skills, Voice Repetition

    lauren (@poteto) shared a Grok Bot team management guide: you don't need many bots—a few good ones are easier to manage than a large bunch; when teaching a bot to do things, have it save the method as a skill so it stays consistent across long conversations. The guide also recommends connecting more connectors, using Vault, and giving bots their own email and accounts; the author most recommends speaking your ideas out loud all at once by voice, then having the bot "repeat them back in its own words" to confirm understanding. Comment @bot under the original post to get a personalized Grok Bot optimization guide.

    Quoted postlauren · @poteto

    here’s a short grok bot guide on how i create high quality bots and manage my team. ask @bot to read this post and help you improve your team! • add dr eggbot to your team! he’s my bot designer that i use to create bots • you don’t need a lot of bots. a few good bots can be more useful and easier to manage than a large team • tell your bots to create skills for itself as you give it instructions on how to do tasks. this lets them do them more consistently even over long conversations. you can even ask it to create a private github repo for you to store all your skills if you want them preserved independently • add connectors, use the vault, and give access to all the tools you use to do your work. you can also ask your bot to setup its own email and it can sign up for its own accounts for better separation • thinking about all the work you do by hand, or steps in your workflows where you have to copy and paste information back to your bot. ask your bot how it can do that for you instead! • imagine you hired an intern, how would you onboard them and help them do the work correctly and efficiently? introduce yourself to your bot and describe the work you do, then ask it how it can help you automate it and my favorite tip: use voice mode or the transcription option! it’s a good way to just brain dump a lot of information. then, at the end, tell your bot to “restate this in your own words” so you can confirm it understood you https://t.co/ajrjeHdzUO

    Same story · 3 reports

    #Tutorials / practice#Agents#English text is being prepared.

  7. Shao Meng
    AI score66

    X and SpaceX design lead open-sources Stageshow faux UI demo Skills

    X and SpaceX design lead Benji Taylor has open-sourced Stageshow skills, used to let agents generate faux UI stills, short animations, and bento grids that look more like real products. Deliverables include SKILL.md and references, a React component kit that can be copied into projects, a large number of gold-standard samples, and a local preview and footage self-check tool film.mjs. The author says the output is ordinary React code rather than unmaintainable images, and states limitations such as checks catching errors but not boring ideas, and most demos still needing several rounds of human feedback.

    Quoted postBenji Taylor · @benjitaylor

    Stageshow: a skill that helps your agent make better faux UI stills, animations, bento grids, and more. npx skills add benjitaylor/skills --skill stageshow https://t.co/Qve3C2cECg

    #English text is being prepared.#Agents#English text is being prepared.

  10. East Money
    AI score40

    Computing power leasing Q3 reports show a tale of two extremes: leading firms' profits multiply severalfold, while small and mid-sized players see revenue growth without profit growth

    The computing power leasing sector's Q3 reports show structural divergence: Xingyun Technology's revenue for the first three quarters was 1.175 billion yuan, up 1896.66% year-on-year; Hongjing Technology's net profit attributable to shareholders was 312 million yuan, up 189.93% year-on-year; while Dawei Technology's revenue grew 28.51% but it posted a loss of 102 million yuan. Contract liabilities became the deciding factor: Dongyangguang's contract liabilities reached 5.779 billion yuan, an increase of 5.683 billion yuan from the beginning of the year, and Litong Electronics saw a year-on-year increase of 4139.94%. The industry is bidding farewell to the era of easy money from "stockpiling cards and subletting" and entering a stage of quality improvement and reshuffling.

    #Industry news#English text is being prepared.

  11. East Money
    AI score56

    S&P Downgrades Meta's Rating Outlook to Negative on AI Spending Pressure

    S&P Global Ratings downgraded the outlook on Meta's AA- credit rating from stable to negative on October 31, citing continued increases in AI-related spending and rising leverage. S&P expects Meta's capital expenditure to reach $140 billion this year and exceed $225 billion in 2027, with leverage rising above 1x as early as next year, and said Meta's ability to convert AI investment into commercial revenue is relatively limited. A Meta spokesperson said the company disagrees with this assessment and believes future product and business opportunities are sufficient to support its credit profile.

    #Industry news#OpenAI#Meta

  12. East Money
    AI score22

    AI boom faces a test: US midterm elections, October nonfarm payrolls, and Apple and AMD earnings arrive next week

    Next week, US stocks will see the US midterm elections, the October nonfarm payrolls report, and earnings from tech giants such as Apple and AMD, as investors reassess AI investment returns and the outlook for monetary policy. This week, the Nasdaq fell 2% cumulatively, the S&P 500 fell 0.57%, and Intel plunged nearly 29% on Friday after its guidance fell short of expectations. The market expects 55,000 new nonfarm jobs in October, with the unemployment rate holding at 4.2%.

    #Industry news#OpenAI

  13. East Money
    AI score74

    OpenAI's target valuation lowered to $1.25 trillion, funding target cut to $20 billion

    OpenAI has lowered the target valuation for this funding round from $1.4 trillion to $1.25 trillion, a reduction of $150 billion, or about 10.7%; the funding target has also been cut from $30 billion to $20 billion, a reduction of $10 billion. People familiar with the matter said that because some investors reduced their subscription amounts, the actual funds raised in the first closing are expected to be less than $20 billion. MGX, part of a consortium of UAE investment institutions, is expected to become the largest participant, while Founders Fund, Coatue, Sequoia Capital, Andreessen Horowitz and SoftBank also plan to participate.

    Same story · 8 reports

    #Industry news#OpenAI

  14. Sina TechProduct release
    AI score66

    DeepSeek Harness Team Releases Super-agent Architecture

    The DeepSeek Harness team published a technical article on its official WeChat account, proposing an Agent form and system architecture called Super-agent. Through designs including a dual-layer architecture, event-driven operation, asynchronous multi-channel interaction, storage-compute separation, and an external Watchdog, it enables the Agent to maintain context across sessions and advance multiple tasks in parallel. The architecture has four basic characteristics: continuous, efficient, proactive, and persistent, and users can interact with the same Agent through interfaces such as IM chats and GitHub comments. The team said that internal instances built on this approach are already running continuously, participating in issue reproduction and fixing, code review responses, and the iteration of DeepSeek Harness itself.

    Same story · 23 reports

    #Product updates#Agents#DeepSeek

  15. Sina TechResearch
    AI score78

    DeepSeek open-sources Super-agent architecture prototype, unveils five-module design

    The DeepSeek Harness team has unveiled the underlying architecture of Super-agent, an autonomous Agent exploration project, and open-sourced the prototype dsh-superagent-prototype on GitHub. The architecture consists of five modules: a two-tier Router/Worker, event-driven design, asynchronous multi-channel interaction, storage-compute separation, and an external Watchdog. Its goal is to enable the Agent to stay continuously online, run multiple tasks in parallel, and proactively respond to environmental events. In preliminary experiments, the p50 task completion time was 12.4 minutes and the p95 was 15.7 minutes, and Router batching reduced the inference time amortized per individual decision from 6.4 seconds to 1.6 seconds. The team stated that the experiments were based on synthetic workloads and a simplified environment, and that this plugin is only a prototype for validating the technical approach, not an official DeepSeek product.

    Same story · 23 reports

    #English text is being prepared.#Agents#DeepSeek

  16. Simon Willison
    AI score65

    Parameter Labs discloses an exploit chain that can take over ChatGPT accounts

    Parameter Labs disclosed an attack chain targeting the ChatGPT HTML preview (formerly Canvas) sandbox. An attacker can craft a share link that, when visited by another logged-in user, steals private data or performs actions on their behalf. The attack starts with the postMessage-receiving code of the MathLive virtual keyboard using innerHTML, injects an iframe without sandbox restrictions, and then uses the YouTube oEmbed endpoint allowed by CSP to reflect and execute JavaScript. OpenAI has fixed the vulnerability and paid a $5,000 bounty.

    #Safety / alignment#OpenAI

  17. Hacker News Trending
    AI score50

    Jiahao Shao Accuses Anthropic of Misjudging His Distillation of Claude, Account Restricted

    Jiahao Shao publicly accused Anthropic and Dario Amodei, stating that he never distilled Claude, that the rollouts came from a strong open-source model, and that Claude was only used to submit tasks and run analyses, yet he was flagged for "training a competitor." He said that operations involving RFT, SFT, or training larger models would result in a ban, that Opus 5.5 was downgraded to Opus 5, or that sessions were terminated outright.

    #Industry news#Anthropic#OpenAI

  18. Karina Nguyen
    AI score26

    Gemini 4 Argon Tops the BallotBench Trading Leaderboard

    Karina says Gemini 4 Argon ranks first on the @ballotbench trading leaderboard. She notes that the realized profits of Argon and fable 5.1 mainly come from repeatedly capturing small price spreads in markets with very narrow win-loss margins; the difference is that Argon never closes positions with stop-losses, thus leaving unrealized inventory risk, whereas fable takes small losses and actively manages exposure. She also says she is bullish on muse.

    #Evaluations / benchmarks#Agents

  20. Shanghai Observer
    AI score30

    2026 World Laureates Forum Opens: Nobel and Fields Medalists Discuss How AI Is Changing the Scientific Research Paradigm

    The 2026 World Laureates Forum opened in Shanghai on October 31, under the theme "Paradigm Shift: The Next Decade of Scientific Breakthroughs." Omar Yaghi, winner of the 2025 Nobel Prize in Chemistry, proposed developing "selection science"; his team used large models to generate 96 sets of candidate crystallization conditions, which were then verified by a robotic platform, and within less than a month the crystallization performance of the material improved significantly. Laurent Lafforgue, winner of the 2002 Fields Medal, in turn asked where mathematicians' creativity lies once machines excel at proof.

    #Industry news#Papers/Research