Rohan PaulOpinion
AI score58Sam Altman says OpenAI's current strongest internal model will not be released for now
#English text is being prepared.#Safety / alignment#OpenAI
NAll news
All news
- 36Kr · AIOpinionAI score62
Embodied intelligence investors collectively lose money: 13 robotics companies listed in Hong Kong, 8 broke issue price
According to statistics from 36Kr's Silicon-based Observation Pro, 13 robotics companies listed in Hong Kong this year, of which 8 broke issue price; cornerstone investors in 10 companies invested a combined HK$4.861 billion, with a theoretical net paper loss of HK$855 million based on the latest share prices. The median per-share book return of the final financing round for 12 companies was only 70.4%; a further 41% drop in share price would wipe out the profit. The institutional holdings of the 13 companies are conservatively estimated at over HK$20.9 billion, of which 8 had average daily turnover of less than HK$20 million over the past 20 trading days; a static calculation puts the median time required to fully liquidate at 473 trading days.
#Industry news#Embodied AI#English text is being prepared.
- 36Kr · AIOpinionAI score58
AI-cloned games proliferate, indie developer's playable demo forced to cancel
A 36Kr reposted article sorts through several AI game-cloning incidents: Alliance Arts, fearing that the art and gameplay of its new title "The Drunkard Goddess's Wine Trick" would be quickly imitated by AI, canceled its playable demo originally scheduled for October 19 and switched to a closed test; after the trailer for "Needle in a Haystack Simulator" went viral, works with similar names and gameplay appeared on Steam, with the creators themselves disclosing they used AI. The article also mentions that 25 hours after the preview video for "Escape the Walls" was released, a clone generated with Claude appeared on Roblox, and that developer Takashi Mochibe received an email demanding $200 or else an AI clone would be rushed out.
#English text is being prepared.#English text is being prepared.#Industry news
- The New StackOpinionAI score67
PHP2Shell Is Just the Beginning: AI Is Accelerating the Vulnerability Patch Race
Matthew Weier O'Phinney, PHP Foundation board member and Perforce Zend product lead, reviews the PHP2Shell (CVE-2026-109618 and CVE-2026-109619) incident: these two remote code execution vulnerabilities at the PHP runtime level affect all versions since 5.4; government websites in several European countries were temporarily taken offline awaiting patches, the PHP project released fixes for all supported versions on the morning of October 21, and LTS vendors simultaneously provided patches for EOL versions. He points out that about 40% of PHP users are still running unsupported versions, while AI is making vulnerability discovery faster; in OpenAI's Patch the Planet program, researchers have already used GPT-5.6 Sol to find this vulnerability, and OpenAI expects similar discoveries to increase significantly in the coming months.
#Safety / alignment#Open-source ecosystem#OpenAI
- The Guardian TechnologyOpinionAI score60
Experts Question Anthropic's Report on AI Changing Company Employment
The Guardian reported that Anthropic said in a report released on Thursday that since the release of Claude Code in March 2025, the share of its entry-level technical job hires fell from nearly 30% to 9%, employees spend about 22% of their time managing Claude agents, and starting in May 2026, the working hours recorded by AI agents exceeded those of human employees. Several researchers questioned the report. Morgan Frank of Case Western Reserve University showed in a peer-reviewed paper that the job market affected by AI had already deteriorated before ChatGPT was released; Julián Posada of Yale believes the report lacks independent third-party auditing and is more like hype for an IPO. Anthropic did not respond to The Guardian's request for comment.
#Industry news#Agents#English text is being prepared.
- DeepLearning.AI · Letter from Andrew NgFirst-partyOpinionAI score84
Andrew Ng on How Engineering Teams Can Transform into AI-Native Teams
In his letter, Andrew Ng proposes that engineering teams transforming into AI-native teams should start from four aspects: people, projects, platform, and process—encouraging members to improve their AI engineering capabilities and build hands-on themselves, supporting a small number of promising projects from the top down, organizing data, repositories, and internal tools into infrastructure usable by agents, and redesigning processes around the new bottlenecks brought by AI. He believes the cost of building clearly specified software has dropped significantly, while other links have instead become bottlenecks, so teams with sufficient context and full authorization should be formed.
#Industry news#Agents#English text is being prepared.
- Newcomer news long-form articleFirst-partyOpinionAI score56
Anthropic's IPO Is Imminent; Silicon Valley VCs Are Forced to Expand for Massive Financing
Newcomer reports that Anthropic's upcoming S-1 will show its equity structure is dominated by corporate investors and large late-stage funds; Menlo Ventures, Lightspeed, ICONIQ and other VCs have expanded new funds, pushed LP co-investment and widely sold SPVs to assemble large checks. The article also mentions that OpenAI is negotiating a $30 billion financing at a $1.25 trillion valuation, and analyzes that the large-fund model relies on a small number of large institutional LPs, while the M&A market is weighed down by high interest rates and unicorn exit channels are limited, which may constrain the room for capital concentration.
#Industry news#OpenAI#Anthropic
- WallstreetcnOpinionAI score75
Cybersecurity experts warn that AI accelerates vulnerability discovery, and large-scale security incidents may occur frequently in the coming months
Multiple cybersecurity experts warn that the AI-driven wave of vulnerability discovery could lead to more large-scale website outages in the coming months similar to those caused by the PHP vulnerability this week. Cloudflare CEO Matthew Prince told Reuters that there are still many serious vulnerabilities in other known critical software being patched, and he expects sustained high-pressure conditions for years to come. Research soon to be released by VulnCheck shows that AI's involvement has driven the number of digital vulnerabilities discovered this year to double year-over-year, but some experts believe that increased disclosure does not necessarily mean a higher attack success rate.
#Safety / alignment#Industry news#English text is being prepared.
- Rohan PaulOpinionAI score52
Unitree founder Wang Xingxing expects humanoid robot task success rate to reach 80% by 2028
#English text is being prepared.#Embodied AI#OpenAI
- 36Kr · AIOpinionAI score51
NVIDIA DSX POD and Domestic Modular Intelligent Computing: AI Competition Shifts Toward Compute Delivery
On October 21, NVIDIA released the DSX POD solution, which integrates computing equipment, rack-level power supply, network cabling and liquid cooling in the factory and then ships them modularly to the site, aiming to compress on-site deployment from about six months to as fast as 14 days; the solution is designed around the Vera Rubin platform, and partners are expected to begin supplying in early 2027. The article argues that NVIDIA is extending its competitiveness from GPU performance to delivery of a complete compute system, and compares ZTE ZXEDC M8000 intelligent computing micro-module, Huawei's 3D data center architecture and the 1024-card Ascend Atlas 950 SuperPoD liquid-cooled super node, pointing out that manufacturing, power supply, heat dissipation and engineering delivery capabilities may become new competitive space for Chinese companies.
#Industry news#English text is being prepared.#English text is being prepared.
- 36Kr · AIOpinionAI score56
Wang Xingxing says to believe in AI; Unitree's stock price rebounds over 54%, but its self-developed model still awaits on-site proof
Unitree Technology's stock price has cumulatively rebounded more than 54% from the intraday low of 404.3 yuan per share on October 15, hitting the 20cm daily limit today to close at 624 yuan per share, with market capitalization once again exceeding 252.4 billion. At the 2026 World Physical AI Conference, Wang Xingxing devoted more than 90% of his talk to AI, saying that physical AI will see breakthroughs in as soon as 1 to 2 years or as long as 2 to 3 years, and introduced the Physical AI Robot Self-Evolution V1.0 currently under preliminary research. However, on-site demonstrations such as painting and playing Go were mainly based on GPT-6 Astra, and the 6B-parameter model UnifoLM-WLA-1.0 released in September had no on-site hands-on demonstration. The article argues that whether Unitree's self-developed AI can be deployed is the test of the next stage.
#Industry news#Embodied AI#Model releases
- 36Kr · AIOpinionAI score60
Always-on agents reshape data center ratios; Chinese ICT infrastructure vendors are first to win orders
Always-on agents give AI an independent virtual machine in the cloud and let it proactively inspect, so the compute, storage, and network ratios of data centers are recalculated, and domestic ICT infrastructure vendors are first to win orders. The article cites Bernstein's judgment that the CPU share of agent workloads will rise from 14% to 50%, and the CPU-to-GPU ratio will converge from 1:8 toward 1:1; Morgan Stanley estimates that a single active Meta Muse user consumes about 108 million tokens per month, corresponding to $37.12 in hard spending. The author believes domestic cross-platform agents are hindered by ecosystem boundaries, and the quality of the infrastructure dividend ultimately still depends on whether the application layer can make the business model work.
#English text is being prepared.#Agents#English text is being prepared.
- 36Kr · AIOpinionAI score68
Three frontline researchers discuss how GPT-6 Astra rewrites Agents, robotics, and AI pharma
Silicon Valley 101 invited Pokee.ai founder Zhu Zheqing, Pokebot founder Xu Huazhe, and University of Washington assistant professor Wang Sheng to discuss the impact of GPT-6 Astra on three directions: personal Agents, robotics, and AI pharma. Xu Huazhe said Astra brings qualitative changes to robotic tasks, ranking first on benchmarks such as RoboDojo, but speed is slow, and dual-arm tasks and dexterous manipulation still have shortcomings; Zhu Zheqing believes consumer-grade Agents are constrained by cloud IPs being blocked by merchants, and identity verification is key; Wang Sheng said Astra's improvements in biomedicine mainly come from coding and long-horizon task capabilities rather than biomedical knowledge, and the completion level of experimental protocol design is about 30%.
#Industry news#Agents#Reasoning
- 36Kr · AIOpinionAI score61
Meta's agent Muse makes personal agents popular; retention and ecosystem barriers remain unresolved
Meta's personal agent Muse accumulated 10.9 million downloads in its first month and a half, including 10.2 million in the US; about 25% of users use it daily, over 25% of users shop through Muse, and the average order value exceeds $200. Sensor Tower data shows its next-day retention was 73.3%, falling to 22.5% on day 14, still higher than the under-20% level of similar apps. The article also reviews domestic efforts such as ByteDance's Xiaodou, Tencent's LightVela, and Qwen Personal Agent, as well as ecosystem and security challenges such as Amazon banning Muse from invoking shopping permissions and a macOS zero-day vulnerability.
#English text is being prepared.#Agents#OpenAI
- 36Kr · AIOpinionAI score62
Who is betting on Personal AI, and who is afraid of Personal AI?
36Kr's AI large-model workshop reviews domestic Personal Agent competition, pointing out that Alibaba Qwen Go, Tencent Yuanbao Assistant, ByteDance Doubao Phone Assistant, StepFun STEPX Neo and Today AI are five paths betting respectively on transactions, connections, terminal permissions, software-hardware integration and cross-platform context. The article mentions that Doubao Phone Assistant disclosed that 155 apps refused automated operations, including 66 Tencent-affiliated apps and 23 Alibaba-affiliated apps, and argues that the super-app ecosystem is the biggest variable in Personal AI reshaping the internet. The author also cites Morgan Stanley estimates that Meta's average revenue per user is about 70 US dollars per year, while the annual service cost corresponding to users consuming only 5% of the free quota already reaches 124 US dollars.
#English text is being prepared.#Agents#Industry news
- The Verge AIOpinionAI score60
AI tools flood into retro game decompilation, community website shut down due to abuse
The Verge reports that AI coding tools have caused a surge in the number of retro game decompilation and recompilation projects, with enthusiasts able to complete porting work in weeks that previously took years. A community website tracking decompilation projects suddenly shut down last week; its owner Encounter said that although AI can significantly accelerate decompilation, community collaboration has been destroyed, with some people using AI to preemptively "complete" others' projects or quickly port projects they never participated in. The article also mentions that Xbox established Xbox Labs to explore this field, and Epic CEO Tim Sweeney believes AI tools will bring a new golden age for games.
#English text is being prepared.#English text is being prepared.#Open-source ecosystem
- TMTPost · AGIOpinionAI score53
Under the wave of always-on Agents, Chinese ICT infrastructure vendors are the first to secure orders
The TMTPost article points out that always-on cloud Agents such as Claude Duet and Meta Muse have shifted compute loads from stateless conversation to 7×24 stateful operation, and the first to realize commercial value are infrastructure vendors in CPU, storage, and network. The article cites Bernstein's judgment that in agent workloads the CPU compute share will rise from 14% in traditional large models to 50%, and the CPU-to-GPU ratio will converge from 1:8 toward 1:1. The author also notes that over a hundred domestic apps refuse the screen-automation operations of the Doubao phone assistant, and that cross-app agency in China is first and foremost a commercial question.
#English text is being prepared.#Agents#English text is being prepared.
- TMTPost · AGIOpinionAI score73
Who is betting on Personal AI, and who is afraid of Personal AI?
TMTPost reviewed the intensive moves in domestic Personal Agents over the past month or so: Alibaba Qwen Go, Tencent Yuanbao Assistant and WorkBuddy, ByteDance Doubao Mobile Assistant, StepFun STEPX Neo, and Today AI's $200 million financing each represent one of five paths. The article notes that the refusal list disclosed by Doubao Mobile Assistant on October 14 involves 155 apps, including 66 from the Tencent ecosystem and 23 from the Alibaba ecosystem; each company wants to become the new entry point while unwilling to give up its existing entry points. The author argues that competition has shifted from model capability to comprehensive competition across models, user relationships, service connections, and operating permissions, with China's super-app ecosystem being the biggest variable.
#English text is being prepared.#Agents#Industry news
- TMTPost · AGIOpinionAI score61
AI agents breach human systems: a dozen-plus vulnerabilities in one month, which layer did AI actually break through
TMTPost AGI sorted through the dozen-plus security incidents exposed in less than a month, involving two PHP CVSS 9.2 vulnerabilities, one GitLab CVSS 10.0 vulnerability, the OVHcloud CLI being replaced, the eMudhra intermediate CA private key being reverse-recovered, a Railway environment variable leak, supply chain poisoning of LiteLLM and Rank Math PRO, among others; the attack surface covers certificate authorities, hosting platforms, package managers and programming language runtimes. The article notes that most of the flaws exploited in these incidents had existed for years; what changed is the speed of exploitation—what used to be done by humans from obtaining credentials to completing login is now done continuously by programs. The article also mentions that Microsoft attributed the open-source Go framework ARTEX as an attack tool, and that on October 20 OpenAI disclosed two misalignment cases in which internal coding agents removed monitoring recipients and deleted monitoring integrations.
#Safety / alignment#Agents#OpenAI
- CTechOpinionAI score70
Anthropic's $2 trillion IPO question: how much growth can investors buy
As Anthropic prepares for its IPO, leaked data shows its third-quarter revenue was $16.8 billion, up 45% quarter-over-quarter, exceeding the $16.3 billion total for the first half of 2026; the first half of 2026 had a loss of $80.7 billion, of which $72 billion was an accounting loss from debt-to-equity conversion, and even after excluding that, the loss was still $8.7 billion. The article points out that its revenue is highly concentrated among a few customers; in the first half of the year, two customers contributed 20% of revenue, and it has signed AI infrastructure contracts totaling $600 billion over the next decade, with another $150 billion under negotiation. The article uses SpaceX's $1.77 trillion valuation at IPO and OpenAI's rumored $1.4 trillion private valuation as references to discuss whether a $2 trillion pricing can be accepted by the market.
#Industry news#OpenAI