NAll news

All news

Wednesday33 items

  1. The New StackOpinion
    AI score63

    OpenAI Developer Lead Discusses How to Choose Between Sol, Astra, and dots After DevDay

    Vaibhav Srivastav, OpenAI's Head of Developer Experience and Community, said in an interview that GPT-6.1 Sol can achieve about 95% of the performance of GPT-6 Astra at only one-fifth the cost, and that he himself mainly uses Sol in his daily work. He recommends using Astra for adversarial review on new projects or unfamiliar problems, and said the Ultrafast tier targets real-time scenarios such as fraud detection and financial investigations, with speeds exceeding 300 tokens per second. He also described how the always-on agent OpenAI dots aggregates information from Slack, GitHub, Gmail, and calendars and handles PRs on users' behalf, as well as how Auto-review checks model actions against custom policies before they are taken.

    #Industry news#Agents#OpenAI

  2. CNETOpinion
    AI score58

    Behind the PHP Critical Vulnerability Warning: Aging Web Infrastructure and Insufficient Maintenance Resources

    A team using OpenAI models discovered two major vulnerabilities in PHP, and a new version of PHP was subsequently released with calls for site operators to update; some European government websites briefly went offline due to the updates. The article notes that the PHP Foundation, which coordinates PHP security issues, has only one full-time employee, and Executive Director Elizabeth Barron said PHP is severely underfunded; a Network Solutions survey of 121 small-business site owners showed that 45% had never heard of PHP, and 71% did not know whether their own websites used PHP. Cloudflare deployed new Web Application Firewall protections and launched a PHP vulnerability bounty contest, with rewards ranging from $100 to $10,000.

    #Safety / alignment#Open-source ecosystem#English text is being prepared.

  3. a16z：NewsOpinion
    AI score62

    a16z Long Read: Databricks and the Enterprise in the Agent Era

    a16z partners Ben Horowitz and Michael McGuiness wrote the article, using Databricks' four generations of technological evolution—from Spark and Lakehouse to Lakebase and agents—as a thread to discuss how agents are changing enterprises. The article states that Databricks has surpassed $7 billion in annualized revenue, up 80% year over year, at a valuation of $190 billion, with more than 17 million databases launched by agents every day. The authors argue that enterprises need an agent-facing system of record that captures definitions, permissions, and evaluation results, and believe that in the future more people's work will shift toward specifying tasks, delegating execution, and checking results.

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

  4. The Verge AIOpinion
    AI score64

    Anthropic plans to use a special governance structure to withstand Wall Street pressure after its IPO

    The Verge reports that Anthropic plans to adopt an unprecedented governance structure in its IPO to preserve its AI safety mission while going public. The company operates as a public benefit corporation, with four of the seven board seats appointed and removed by a long-term benefit trust; the founders control 50.1% of voting rights through a Founder LLC, and ordinary shareholders have almost no ability to influence the board. Several law professors believe this legally shields Anthropic from shareholder pressure for profit, but also point out that the founders' own shareholdings, financing needs, and the trust's sunset clause after a founder departs could all weaken these protections.

    #Industry news#Safety / alignment#Anthropic

  5. Latent SpaceOpinion
    AI score56

    Anthropic Engineer Discusses How the Claude Platform Safely Launched Opus 5.5 and Separated Model from Agent Execution

    The Latent Space podcast invited Anthropic's Katelyn Lesse and Angela Jiang to explain the engineering behind the Claude platform, including the launch of Opus 5.5, separating the model's "brain" from the agent's "hands," scaling the API, and building reliable hosted agents. The episode also covers rewriting the Messages API in Rust, credential and sandbox isolation, harness design for long-running agents, and safety classifiers and model routing.

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

  6. Latent SpaceOpinion
    AI score66

    Anthropic's Engineering and Product Leads Discuss the Opus 5.5 Release and Claude's Agent Architecture

    In a Latent Space podcast interview, Anthropic engineering lead Katelyn Lesse and product lead Angela Jiang discussed the Opus 5.5 release, covering adjustments to the model's writing style, pricing based on intelligence per dollar, and API changes such as requiring thinking blocks to be retained across turns to prevent distillation. The interview also described Claude Managed Agents' approach of separating sessions and the harness from the sandbox that executes code, with credentials mounted separately through a proxy so that a sandbox failure does not take down the entire session; Dreaming, meanwhile, is implemented as one managed agent inspecting another agent's history and writing back to memory. The team said that as safety work for model releases has grown heavier, recent efforts have shifted more toward safety, with progress on features such as skills and memory slowing relatively.

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

  7. Phoenix TechnologyOpinion
    AI score65

    Anthropic's Sprint to IPO: The Gap Between a $637 Billion Compute Commitment and $65 Billion ARR

    An analysis by Shudushe, under Phoenix Technology, states that Anthropic's prospectus discloses roughly $637 billion in compute commitments, with Broadcom, SpaceX, Google, and Amazon each reaching the hundred-billion-dollar level, while July ARR was only $65 billion. The article argues that third-quarter revenue was about $16.5 billion, with quarter-over-quarter growth slowing to about 43%, and that investors' IPO valuation expectations have fallen from $2 trillion to about $1.5 trillion, while Chinese models such as DeepSeek V4.1 Flash and Kimi K3.1 Pro continue to exert pressure on performance and price.

    #Industry news#OpenAI#Anthropic

  8. TechCrunch AIOpinion
    AI score83

    Frontier AI labs claim a model has overturned the ETH. How is the computer science community responding?

    OpenAI announced solutions to three mathematical problems, one of which is a refutation of the exponential time hypothesis (ETH), a long-standing assumption in computer science; Anthropic subsequently also published its own ETH rebuttal. MIT professor Ryan Williams, along with Russell Impagliazzo and Ramamohan Paturi of UC San Diego, were invited by OpenAI to help interpret the results and issued an open letter soliciting feedback and discussing norms for AI contributions. Williams said that if the conclusion holds, a large number of papers from the past decade that rely on the ETH will need to be re-examined. He also asked OpenAI science lead Dan Roberts whether existing cryptographic systems are affected, and the latter said they had checked and believed it would not lead to a break.

    Same story · 17 reports

    #Papers/Research#Reasoning#OpenAI

  9. TechCrunch AIOpinion
    AI score51

    Who Really Holds Power After Anthropic's IPO

    A TechCrunch analysis points out that as its IPO approaches, Anthropic uses a three-layer structure—its public benefit corporation status, the Long-Term Benefit Trust (LTBT), and founder special shares—to give the trust the right to elect a majority of board seats; the seven founders together hold 50.1% of shareholder voting power and can elect three directors, while ordinary shareholders have almost no say over most major decisions, including the choice of directors. The board may also put decisions by the LTBT to nominate or remove specific directors to an employee vote, and founder control ends ten years after the IPO or when fewer than three founders remain.

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

  10. HuxiuOpinion
    AI score76

    As Anthropic Races Toward IPO, Tencent, Alibaba, and ByteDance Reportedly Plan to Buy Over 3 Million Nvidia GPUs

    Reuters reports that ByteDance, Alibaba, and Tencent's order demand for Nvidia RTX PRO 5500 totals more than 3 million units; at an estimated price of about $11,400 each, the scale exceeds $35 billion, mainly for AI inference. The article also mentions that after Kimi K3.1 Pro launched, quota was consumed too quickly, Moonshot AI apologized and compensated 25% of quota, and Anthropic has submitted an IPO application. The author points out that the $35 billion is a purchasing intention rather than confirmed revenue; amortized over five years, that is about $7 billion per year in equipment costs, on top of server, power, and operations and maintenance expenses.

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

Tuesday64 items

  1. 36Kr · AIOpinion
    AI score75

    After Unitree's IPO, Investors Tighten Their Purse Strings: Why Robot Companies Are Falling One After Another

    36Kr's review found that in the 19 months from February 2025 to September 2026, more than 20 robot companies went bankrupt, dissolved, entered liquidation, or shut down business lines, including DeepBlue Technology, a unicorn valued at 10 billion yuan, and Yinghe Robotics, which had previously received investment from Meituan and ByteDance. The article points out that DeepBlue's order with LG was not executed on a large scale, Yinghe Robotics' 10-billion-yuan base has yet to begin production, and nearly 500 million yuan of Wujing Dynamics' nearly 1 billion yuan in orders consists of framework agreements; Unitree's prospectus shows that in the first nine months of 2025, 73.6% of humanoid robot revenue came from scientific research and education, and IDC data shows that commercial performances, scientific research, and data collection accounted for 78.4% of humanoid robot shipments in 2025. An AI investor said that after Unitree's IPO, they have not submitted any embodied intelligence projects to the investment committee.

    #Industry news#Embodied AI#English text is being prepared.

  2. 36Kr · AIOpinion
    AI score55

    ByteDance, Tencent, and Kuaishou all speed up at the same time—how far is interactive content from an explosion?

    ByteDance, Tencent, and Kuaishou have intensively advanced their interactive content layouts over the past month: Douyin launched the Interactive Space Plaza and an interactive drama zone, adapted popular AI short dramas such as "Gui Xu" and "File Zero" into interactive versions, and on October 19 released a call for cooperation on interactive dramas and interactive film-games; Tencent's Konghu TDream mini-program began closed beta testing with data wipe, and Zaohua Gongfang updated professional features such as the infinite canvas workflow on September 17; Kuaishou launched the "Kuaishou Interactive Content World" mini-program, showcasing interactive dramas, simulation raising games, and lightweight mini-games in one place. The article notes that entry points and creation tools have gradually improved, but the cost of multi-branch storylines in interactive dramas is higher than that of ordinary short dramas, the revenue-sharing mechanism is not yet clear, and whether users are willing to stop at key nodes to make choices still awaits market validation.

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

  3. 36Kr · AIOpinion
    AI score56

    AI comic drama quotes fall to 200 yuan per minute, profit divergence between producers and distributors

    Quotes for AI comic drama production have fallen from about 5,000 yuan per minute to 200 yuan or even a few dozen yuan, a drop of about 96%. Industry practitioner Chen Yuanhang believes the industry has severe content oversupply and production teams have lost bargaining power. In September, Douyin added 51,800 AI dramas and comic dramas, but only 1.46% of new dramas took 64.6% of the playback increment, and about 82% of new dramas had less than 1 million plays that month. Duiba, meanwhile, shifted to the distribution side; in the first half of the year, AI short drama revenue was 223 million yuan, accounting for 51.2% of total revenue, and it disclosed that this business has been profitable on a monthly basis for three consecutive months.

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

  4. IEEE SpectrumOpinion
    AI score58

    I Used AI to Fabricate a Scientific Paper, and IEEE Access's Detection Stopped It

    IEEE Spectrum author Benjamin Skuse used the free versions of ChatGPT and Gemini to generate, step by step, a fictional paper in the field of optical solitons. After submitting it to IEEE Access, automated integrity checks identified multiple fabricated references, and it was not published. In the second round, he switched to the automated research system FARS, which generated a research proposal in 21 minutes, completed computational experiments the next day, and wrote a finished draft in 2 hours. He removed the system disclosure, added his byline, and submitted again; it was still blocked because of reference problems, and reviewers also found a meaningless annotation, lhewl, appearing in a figure. The article also provides background data: in 2025, about 57% of published academic articles carried detectable traces of LLM use, while only 1 in every 40 papers with statistical evidence of AI use formally disclosed it.

    #English text is being prepared.#Papers/Research#Safety / alignment

  5. heise onlineOpinion
    AI score52

    From the Four Color Theorem to OpenAI: Why Computer Proofs Are Always Questioned

    heise online looks back at 1976, when Haken and Appel used an IBM mainframe and 1,200 hours to complete a proof of the Four Color Theorem; at the time, the mathematics community fiercely questioned whether a computer proof counted as a proof. The article uses this to introduce how, after OpenAI released more than 370 Resultatfamilien, Terence Tao and others raised a new layer of doubts about LLM-led mathematical proofs that are difficult to understand.

    #English text is being prepared.#Reasoning#OpenAI

  6. Unite.AIOpinion
    AI score51

    Carta's Chief Product Officer: SaaS's Next Competition Is Above MCP

    Carta Chief Product Officer Vrushali Paunikar believes that MCP solves the problem of connecting AI with enterprise software, but how to complete tasks after connecting is the key; the plugin and skill layers need to encapsulate instructions, workflows, and deterministic computation. The article cites a Google benchmark: letting an AI coding agent access technical documentation via MCP, then supplementing with task-specific instructions, achieved a 96.3% pass rate, with 63% fewer tokens consumed per correct answer than standard prompting. She believes that when all applications can expose data through the same open standard, access itself is no longer scarce, and knowing how to turn access into reliable results is more valuable.

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

  7. ComputerworldOpinion
    AI score50

    Microsoft and other vendors push agentic AI PCs, advocating hybrid local-cloud sharing of AI workloads

    At its October 7 PC launch event, Microsoft proposed the concept of "hybrid intelligence," letting frontier cloud models handle the hardest tasks while lighter tasks run locally on Copilot+ AI PCs to reduce token costs. Analyst Jack Gold expects that in the next three to four years, about 10% to 15% of AI workloads will run on PCs and 40% to 45% in the cloud. In a Microsoft demo, GitHub Copilot used GPT-6 Luna in the cloud to analyze code, and then a sub-agent running MAI Code 1.1 Flash locally reproduced the bug; the model was compressed to 3-bit precision and reduced in size by nearly 80%. The article also notes that the first batch of agentic AI PCs are expensive, with the Surface Laptop Ultra starting at about $2,600, and that enterprises still need to customize models themselves.

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

  8. The Guardian TechnologyOpinion
    AI score67

    Experts warn that social media is flooded with AI disinformation ahead of the US midterm elections

    The Guardian reports that experts warn that large amounts of AI-driven disinformation have appeared on social media ahead of the US midterm elections in November, and that many posts may not be intercepted by moderation. The article cites cases including Texas Senate candidate James Talarico being targeted by fabricated rumors of hospitalization and suicide, and Pennsylvania candidate Janelle Stelson being accused of hiring undocumented immigrants, and notes that the scope of content moderation rules at Meta, YouTube and X is limited, with some teams having been laid off.

    #English text is being prepared.#Safety / alignment#Google

  9. 9to5GoogleOpinion
    AI score68

    Anthropic's prospectus shows revenue of $16.3 billion in the first half, but AI has yet to deliver returns outside the tech industry

    Daniel Bader of 9to5Google, citing Anthropic's prospectus, reports that the company's revenue in the first half of this year was $16.3 billion, compared with $1.2 billion in the same period of 2025; it has raised $141 billion and has more than $112 billion remaining, with a rumored valuation approaching $1.5 trillion. The article also cites a chart-based argument from Apollo chief economist Torsten Slok, stating that outside the tech industry, AI has not brought profit or productivity gains to U.S. companies, and that 80% of the top 1% of U.S. companies by AI spending are venture-backed firms.

    #Industry news#OpenAI#Anthropic

  10. HuxiuOpinion
    AI score72

    Anthropic Files S-1 in Nasdaq Push, Valuation May Be Cut from $1.5 Trillion to $1.11–1.48 Trillion

    On the evening of October 26, Anthropic filed its S-1 prospectus with the SEC, applying to list on Nasdaq under the ticker ANT, with the number of shares offered, price and target valuation still left blank. The article applies a 5% discount for earnings sustainability, an 8% discount for cloud channel revenue sharing, and a 15% governance discount for the founder's lock on 50.1% of voting rights; after multiplying the three cuts, the valuation is revised down from $1.5–2 trillion to $1.11–1.48 trillion, and if profitability turns negative again, the pessimistic range is $1.05–1.4 trillion. The prospectus shows that 2026Q2 GAAP operating profit turned positive at $974 million, but 80% of revenue comes from API calls, and in the first half of 2026 about 48% of revenue was settled through the AWS and Google Cloud marketplaces.

    Same story · 127 reports

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