NAll news

All news

Thursday55 items

  1. Shao MengOpinion
    AI score60

    Nathan Lambert Harvard lecture: Chinese open-source models are fully ahead; US compute advantage is about 16x

    In a guest lecture for Harvard CS 2881R "AI Security," Nathan Lambert judged that open-source models lag the closed-source frontier by about 3 to 5 months and the gap is narrowing; Chinese open-source models are fully ahead in performance, downloads, and paper citations, while US open-source models lag China by another 3 to 4 months. He believes treating distillation as a core issue overestimates it; the US's truly durable advantage is compute, with the combined compute of the top five labs averaging 16 times that of China's top five, and by the end of 2026 Anthropic and OpenAI each having more than 5 GW of compute, while China's leading labs have at most 100 to 200 MW. He also pointed out that banning open source is not feasible in reality and would only suppress economic activity, and suggested assessing additional risk on a per-model basis with unified national standards.

    Quoted postNathan Lambert · @natolambert

    Finally got to put together a complete deck on the state of open models. What they are, licenses, how competitive they are, signs people are using them, distillation's impact, and how we balance the risks and benefits of openness! Will record a video lecture of these soon, and hopefully re-do them a few times a year.

    #English text is being prepared.#Open-source ecosystem#English text is being prepared.

  2. Shao MengOpinion
    AI score67

    Claude Mods In-Depth: Harness Shifts from Managing Machines to Managing Human Attention

    Shao Meng relays Dan McAteer's article on latentspacepod, arguing that Claude Mods lets users and Claude work together to use code to transform the human-facing layer of Claude Code—deciding when to interrupt, what to display, and which decisions to let users make—and that it is an early form of an attention interface. Mods are essentially TS/JS plugins that listen to prompt, tool-call, and interface-rendering events, and work in both the CLI and the desktop app. The official release already includes 6 built-in Mods: agents-md, diff, plugin-authoring, sec-default, telemetry, and you-should-know. Dan also points out the noise problem: after installing it on October 15, Supervisor was triggered in 99 out of 111 rounds; in addition, each check consumes about 5–6k equivalent input tokens, and Mods run with user permissions and have no sandbox.

    Quoted postDan McAteer · @daniel_mac8

    There's not enough buzz around Claude Mods. Read this Claude Mods article if you want to: - Understand Claude Mods basics - Install and use the most useful mods today - Ponder where the agent-human interface will go For @latentspacepod.

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

  3. Alex HeathOpinion
    AI score58

    Sundar Pichai responds to Google's free cash flow turning negative for the first time while still ramping up AI capital expenditure

    Alex Heath asked Google CEO Sundar Pichai why the company recently let free cash flow turn negative for the first time in order to increase AI capital expenditure, and Pichai responded that it is not based on hope alone. He said making this decision requires confidence on multiple levels: Google sees early signals of diverse opportunities and strong demand from consumers to enterprises; at the same time, the company itself, as an AI user, also has demand signals and can independently assess progress at the capability frontier, and research breakthroughs led him to judge that Gemini 5.0 will be significantly stronger than Argon. He also mentioned that Google is signing long-term contracts with committed revenue, so it is planning its business portfolio and investing in a responsible and disciplined manner.

    Quoted postAlex Heath · @alexeheath

    Sundar Pichai on Gemini 4 and the AI compute crunch Even Google can’t get enough compute. I sit down with Google and Alphabet CEO @sundarpichai shortly after the wider release of Gemini 4 Argon to discuss its early reception, where he still wants the model to improve, and what it takes to keep Google at the frontier of AI. Pichai explains the recent leadership shakeup at Google DeepMind and why building frontier models has become an “all-consuming” job. We also get into Google’s answer to the personal agent race, how he predicts Gemini will do more work on your behalf, and whether agents will finally bring Search, Gmail, and Gemini together. I ask him about Google turning free cash flow negative, the risks in the AI infrastructure buildout, and how he plans for demand years before knowing what the models will be capable of. He explains the signals giving him confidence to keep spending and how he uses AI agents to follow model development and compute usage inside Google. Timestamps: 00:00 Gemini Argon launch reception 01:49 Model access and safety 06:20 The AI model race 08:35 Agentic breakthroughs and 2027 11:34 Recursive self-improvement 15:41 DeepMind leadership changes 19:52 Google's AI compute investments 25:20 Is AI spending sustainable? 33:18 How Sundar uses AI agents 35:35 Gemini's personal agent plans 38:37 How AI agents change the web 43:52 Google's space data centers 47:38 Waymo's expansion strategy 50:26 Wing's drone delivery plans 51:36 AI safety and regulation 55:58 Google's quantum computing bet YouTube: https://t.co/1Cc25LitmB Spotify: https://t.co/XnD4pEEsh8 Substack: https://t.co/0SrnNGKqlm Apple Podcasts: https://t.co/8UKvyXTlFF This episode's presenting sponsor is @hyperagentapp. Thanks to the show’s premiere sponsors: @Atlassian, @meetgranola, and @mercury.

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

  4. Thomas WolfOpinion
    AI score61

    Thomas Wolf comments on Recursive's new model Flint: quantizing lm_head to 4 bits still comes close to bf16

    Hugging Face co-founder Thomas Wolf said that his AGI standard remains AI winning a Nobel Prize, but the AI researchers at Recursive quantized lm_head—which is usually advised to be kept at high precision—to 4 bits, and the result differs from bf16 by only 0.7 percentage points, which impressed him. He also mentioned that the model was released as open weights on Hugging Face, and congratulated Recursive. The quoted original Recursive post introduced that Flint is its first model, positioned as a fast local coding model, built by the automated research system Eureka Machine, reaching 125+ tokens/sec on a MacBook Pro and 150+ on RTX Spark, and built from scratch in under 4 months.

    Quoted postRichard Socher · @RichardSocher

    Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

    #English text is being prepared.#Open-source ecosystem#English text is being prepared.

  5. Ethan MollickOpinion
    AI score70

    Ethan Mollick: AI attack-defense acceleration is creating a gap with human response systems

    Ethan Mollick reposted a post summarizing multiple security incidents of the day, calling it a visible sign of the phenomenon that people in the AI circle have been warning about since the release of Mythos. He believes that attack and defense are accelerating at non-human speed and interacting with human systems that need to respond. The reposted post listed events including two critical PHP vulnerabilities that allow unauthenticated remote code execution, a GitLab vulnerability with a CVSS score of 10.0, LiteLLM 1.107.0 being hijacked to implant a credential stealer and a remote access trojan, Railway suspending all deployments, and TiDB Cloud being breached.

    Quoted post⁵⁄₉ · @fiveoutofnine

    So basically, today PHP got two critical bugs that let anyone run code on your server without even logging in, which is bad, and then the details leaked before the patch was even out, which is worse, and it works on WordPress and MediaWiki and Magento and Joomla and Laravel, and Censys says like 1.5 million servers are running a version with no fix, so the NSA was like "take your system offline," and the FBI was like "take your system offline," and CISA and the UK and CERT-EU were all like "take your system offline," and then the Prime Minister of Japan, the actual Prime Minister, posted at 5:55 in the morning telling 48.8 million people to take their system offline, so everyone took their system offline, like the Dutch cybersecurity agency took its own website offline, and the Dutch parliament went offline, and the Port of Rotterdam went offline, and Finland and France and Denmark started pulling government sites, and Wikipedia went down, and https://t.co/QlOXb3COq3 went down, and Asana and GOG and Something Awful went down, and the University of Vienna's webmail went down for some reason, and that's just PHP. Because GitLab also dropped a CVSS 10.0, which is the highest score they give out, for a bug that's been there since 2014, so that's every self-hosted GitLab for twelve years, and then LiteLLM 1.107.0 got hijacked and shipped a credential stealer and a remote access trojan in every single wheel and the Docker images, including latest, and you didn't even have to import it, just installing it was enough, and it's the second time this year, and then Railway paused every deployment on the entire platform because someone got into the environment variables, so now everyone has to rotate every secret they've ever stored there, and TiDB Cloud said someone was walking around inside with stolen credentials for basically a month and an extortion group says they took 10.6 terabytes, and Zeabur got breached for the second time in two months and is just shutting down, like the whole company, and Node.js patched remote code execution and a TLS bypass, and CISA says Plex is being actively exploited, and someone took over 300 Tor exit relays to steal SSH passwords, and SafePay ransomware has hit over 600 critical infrastructure orgs, and India's biggest bank filed a data loss incident, and Japan has been in an official state of emergency in cyberspace since October 9, which honestly explains the 5:55 AM post. Oh, and eMudhra, a certificate authority that's issued over 60 million certificates, whose entire job is proving websites are who they say they are, was supposed to be using private keys that are random numbers over a hundred digits long, and one of them was 4. Just 4. And the others were like 66 and 237. Anyway, it's Wednesday.

    #English text is being prepared.#Safety / alignment

  6. YicaiOpinion
    AI score55

    Wu Yue targeted by AI face-swap livestream selling: account ban came after public outcry—why does governance always lag behind?

    Actress Wu Yue had her likeness counterfeited through AI face-swapping and voice synthesis for livestream selling; the account involved was permanently banned by the platform after she spoke out and the media reported on it—the ban came after the infringement rather than at the time it occurred. The author, Fan Zhongchen, argues that generating a face-swap video costs only a few dozen yuan, while victims face months of evidence collection, complaints, and litigation, creating a severe asymmetry between gains and costs. The article proposes that platforms should intercept in advance at every stage—registration, going live, distribution, and monetization—especially by binding and verifying three parties: the person appearing on camera, the account, and the payment recipient, and by using livestream sales revenue as the base for damages.

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

  7. CNN Business TechOpinion
    AI score69

    AI Writing Detection Tools Spark Controversy in Literary World: Pangram False Positives and Allegation Bias

    CNN reviewed multiple AI writing controversies in literary awards in 2026, including Caine Prize winner Simbarashe Kundizeza being flagged as AI-generated by Pangram and Goncourt Prize finalist Thélyson Orélien being disqualified over detection allegations. Pangram claims a false positive rate of one in ten thousand, but a 2025 University of Maryland study found that it mislabeled 2% of documents as AI; multiple controversies have centered on non-white writers. The Commonwealth Foundation chose not to rely on a single detection tool, instead making its judgment based on drafts, notes, and interviews, and ultimately determined that the winner did not use AI.

    #English text is being prepared.#Safety / alignment

  8. Phoenix TechnologyOpinion
    AI score51

    Agent phones poised for explosive growth: the upgrade dilemma, on-device AI, and the battle for the next-generation entry point

    Silicon Valley 101 sorts out the drivers and diverging approaches behind the accelerated rollout of agent phones in 2026: the replacement cycle is moving from roughly 43 months in 2023–2024 toward 48 months, manufacturers need new reasons for users to upgrade, and agents turn the phone from a tool that waits for people to operate it into an assistant that gets things done on their behalf. The article compares two approaches—the whole-device agent and the SuperApp—the former represented by the Nubia NaviX Ultra with the Doubao phone assistant, and the latter by entry points such as ChatGPT, Doubao, and Gemini that call on various services, with Apple, Google, and OpenAI each making different trade-offs. The article also draws on the iteration of Qualcomm's Snapdragon platforms and new products such as the Snapdragon 8 Elite Gen 6 to explain that low-power on-device sensing, heterogeneous scheduling, and device-cloud collaboration are the key conditions for agents to run.

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

  9. Phoenix TechnologyOpinion
    AI score64

    Anthropic CEO Dario Responds to Safety Concerns in a Long Post, Reveals High-End Model Was Held Back for Four Months

    Anthropic CEO Dario posted a long message on X responding to questions from AI safety investigator Nathan Calvin, disclosing for the first time that the company had already trained a high-end model at the Mythos level as early as February this year. Because its cybersecurity defense mechanisms were not yet complete, it was released only to a very small number of partners, and was not opened to the public until the safety guardrails were in place in June—about four months in total. He also listed the costs the company has proactively borne for safety, including a large number of false positives caused by biological risk filtering, pushing for legislation such as California SB 53, and a plan to commit 5% of the company's shares to a public interest fund. Dario said this approach amounts to competing toward a higher standard, while Calvin questioned how much more the company can concede in the race for the sake of safety.

    Same story · 3 reports

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

  10. WallstreetcnOpinion
    AI score62

    Personal Agents May Spark a Battle to Defend Bank Deposits

    A Wallstreetcn analysis argues that once personal agents obtain account authorization, they may identify idle funds for customers, compare yields, and transfer funds out, dealing a blow to the low-cost demand deposits that banks rely on. The article notes that when Doubao's phone assistant opened screen automation operations, 151 apps including ICBC and China Merchants Bank explicitly refused, while Citi, Bank of America, and JPMorgan Chase also tightened external agents' access to account permissions. The article uses Yu'ebao as a reference and cites a scenario estimate by Apollo chief economist Torsten Slok, namely that household cash could shift from demand accounts yielding about 0.1% to accounts yielding 3.3% to 5%.

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

  11. 36Kr · AIOpinion
    AI score62

    AI can revive old games in just 8 hours, but the era of mass revival may be short-lived

    A player used AI to revive Com2us's "Magic War" in 7 hours and 53 minutes, consuming about 1.08 billion tokens and sending only 52 messages, without writing any code throughout. In contrast, the team that turned an old version of "Asphalt 8" into a standalone offline version spent more than 6 years, and only released a relatively polished version on October 3. The article also mentions that revived games are being resold at low prices by second-hand scalpers, which may prompt manufacturers to tighten access to resources; on October 16, the reverse-engineering community decomp.dev announced its shutdown.

    #English text is being prepared.#Agents#Open-source ecosystem

  12. 36Kr · AIOpinion
    AI score57

    QuestMobile Q3 Report: Doubao Tops with 419 Million MAU, Tencent WorkBuddy Leads Office Agents

    QuestMobile released its Q3 2026 AI application market report. In September, Doubao ranked first among AI-native apps with 419 million MAU, while Qwen and DeepSeek placed second and third with 158 million and 123 million respectively. On the office Agent ranking, Tencent WorkBuddy led with 32.353 million MAU, followed by Baidu Dazi and Qwen Office in second and third with 9.77 million and 7.905 million. The report shows that in September, the AI office sector reached 142 million MAU, up 50.6% year-over-year, while PC client users of office Agents grew 512.2% year-over-year. The article compares the different layouts of ByteDance, Tencent, Alibaba, and Baidu across general-purpose entry points and office scenarios, and cites analysts' views that competition is shifting from traffic entry points to task entry points.

    Same story · 2 reports

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

  13. ZhidongxiOpinion
    AI score65

    Personal agents such as Muse and Doubao are banned by platforms, and cross-site tasks hit roadblocks

    Zhidongxi reviewed cases in which personal agents such as Meta Muse, xAI Grok, and OpenAI Dots were restricted from accessing platforms including Amazon, Yelp, eBay, Delta Air Lines, and JPMorgan Chase. Domestically, the Doubao mobile assistant also disclosed on October 14 that 155 apps, including WeChat, Taobao, and Alipay, refused its automated operations. The article reviews the lawsuit between Amazon and Perplexity; in August, an appellate court overturned a ruling that required Perplexity to stop agent-based shopping, and the legal boundary remains unsettled. Meta instead partnered with retailers such as Walmart, Gap, Sephora, and Best Buy to enable shopping, while Alibaba's Qwen Go prioritized completing errands within its own ecosystem of Taobao and Alipay.

    #English text is being prepared.#Agents#OpenAI

  15. Tomasz Tunguz blogOpinion
    AI score67

    GPU revenue ladder: from DeepSeek's $38 to SpaceX's $2.44

    Tomasz Tunguz converts the inference data disclosed by DeepSeek and SpaceX's third-quarter financial report into the same basis, yielding four price points for revenue per GPU card-hour, from DeepSeek's theoretical maximum of $38 to SpaceX's actual $2.44. The B200 spot price is $8.77, the SpaceX contract price is $5.27, and the payback period for its new compute deployment is less than one year. The author also points out reverse pressure: GPU rent rose from $4.40 to $8.08 within half a year, while the cost of the equivalent benchmark threshold fell 377 times in 18 months, and JPMorgan says more than 60% of the data center capacity planned for 2027 has not yet started construction.

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

Wednesday47 items

  2. SolidotOpinion
    AI score60

    AI-generated content increases workload for UK civil servants

    UK government departments are being flooded with AI-generated freedom of information requests, letters from members of Parliament, and legal complaints, and staff must read and screen each one manually. Labor arbitration cases in England, Scotland, and Wales surged 39% within a year, partly because plaintiffs are using AI to draft complaints, leaving judges overwhelmed by cases riddled with flaws. A special adviser to the UK government said departments are being inundated with AI-generated freedom of information requests, and if the public widely adopts more automated AI agents, the risk of government agencies becoming paralyzed will further intensify.

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