Models, products and AI startups, near-duplicates collapsed to the most credible source
Researchers urge top conferences to reject AI papersA poster reports cases at top AI conferences where authors could not explain their own work, answered questions live via ChatGPT, faked data, or double-submitted the same paper, and urges conferences to desk-reject AI-generated papers, arguing such behavior crowds out genuine researchers and harms trust in science.
NICE to cut about 500 jobs, roughly 5% of workforceIsraeli software firm NICE is cutting about 500 jobs, roughly 5% of its 10,000-strong global workforce, with its ~1,000 Israeli staff relatively protected. The move follows stock declines and pressure from AI agents on customer-service software; NICE is also weighing a sale of its Actimize fraud unit.
Axion 1.0 open agent foundation model releasedAxion 1.0 is a 35B fully open agent foundation model trained with only two supervised fine-tuning stages plus reinforcement learning, skipping mid-training, DPO, and OPD. The team released the full recipe including data construction, verification, prompts, and training code, claiming competitiveness with leading commercial models on several agentic benchmarks.
Hydrogen-1.0 open-weight science model releasedDeep Principle released Hydrogen-1.0, an open-weight model post-trained from GLM-5.3 for chemistry, materials and life sciences. It scored 38.3% on FrontierScience Research versus 33.3% for GPT-6 Astra, at an estimated ~72% lower output cost, and 92.9% on GPQA Diamond versus Kimi K3's 93.5%.
HDI Seguros and insureMO use AI agents for insurance product engineeringBrazilian insurer HDI Seguros and insurance platform insureMO are expanding their partnership to apply generative AI agents to product engineering. After migrating 11 million records and launching 19 products in under nine months, with premiums exceeding R$500 million (~$93.5M), they say AI agents can cut development and configuration time by up to 85%.
Google to spend $15 billion expanding Malaysia data centersAlphabet's Google plans to invest $15 billion to expand data center and cloud operations in Malaysia, citing AI-driven demand for digital infrastructure. The move reflects Big Tech's continued push to build AI capacity in Southeast Asia.
Nvidia DGX Spark encrypted firmware cuts Linux inference speedOn Linux, Nvidia's DGX Spark laptop locks the GPU to about 780 MHz versus 2.2 GHz on Windows, because its encrypted power firmware is only readable by the Windows driver. GPT-OSS-120B inference drops from 65.4 to 13.9 tokens/s. WSL keeps full performance; a real fix needs Nvidia.
Meta debuts euro bond sale after $15 billion US dealMeta Platforms is entering the euro bond market for the first time, after a $15 billion US bond sale. The move joins a borrowing wave by major tech companies racing to finance trillions of dollars in AI investment.
OpenAI pipeline finds 948 Linux kernel vulnerabilitiesOpenAI built an autonomous pipeline that found, validated and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel, showing AI's ability to automate vulnerability discovery and remediation.
Community teases open-weights model Hy4A post on Reddit's LocalLLaMA community teased a new model called Hy4, saying its open weights are coming soon. No details on size or performance were disclosed.
AI stocks slide as Texas governor result raises data center risksThe Texas governor election result has heightened concerns over data center policy, sending AI-related stocks lower. The Hacker News discussion reflects growing investor sensitivity to political and regulatory risks facing AI infrastructure investment.
Daybreak Red launches in AWS GovCloudDaybreak Red gives approved customers GPT-5.6 Cyber in AWS GovCloud (US-East) via Amazon Bedrock for authorized cybersecurity work, widening access to a frontier model for US government and compliance-bound users.
Visa delays frustrate NeurIPS Sydney attendeesReddit users report visa processing for NeurIPS Sydney attendees takes nearly a month, with rejections despite invitation letters and complete documents, and organizers unresponsive when visas fail. The issue highlights travel barriers for international academic conferences.
China robot funding slows to $1.6 billion as shakeout fears growFundraising in China's robotics sector fell to about 11 billion yuan ($1.6 billion) in October, a sharp cooling from the earlier frenzy. Investment momentum has eased since Unitree's Spring Festival humanoid showcase, with fears of an industry shakeout growing.
Democratic sweep rattles Trump trades from AI stocks to cryptoA Democratic sweep of the US Congress jolted global markets, pushing down Asian tech stocks and the dollar while Treasuries gained. Investors expect greater oversight and investigations of Trump's spending plans, forcing a rethink of trades built on deregulation and rapid AI infrastructure buildout.
Tencent releases foundation model, says it ranks among China's bestTencent released a foundation model on Wednesday that it said joins the top tier of Chinese open-weight AI systems for the first time. The Bloomberg report did not disclose the model's name or other details.
USI unveils Wi-Fi 8 roadmap targeting AI devicesUSI announced its Wi-Fi 8 (IEEE 802.11bn) roadmap with products planned for Q1 2027, shifting focus from peak data rates to ultra-high reliability, efficiency and consistent experience. The standard adds multi-AP coordination, dynamic spectrum optimization and PHY enhancements for small AI devices such as smart glasses, smart rings and TWS earbuds.
ADIPEC 2026 opens in Abu Dhabi to address AI energy demandADIPEC 2026 opened in Abu Dhabi with an expected 239,000 participants from over 170 countries, the event's largest edition. Attendees including OpenAI CEO Sam Altman examine energy resilience, infrastructure investment and AI's impact on energy security; ADNOC's Al Jaber urged rethinking how secure energy is built and delivered.
Suffolk Invests $50M in Boston Construction Innovation Hub 100MAGUS builder Suffolk opened 100MAG, a $50M facility at its Boston headquarters campus, combining AI, robotics, quantum and data technologies to test and scale solutions across the building lifecycle. The company cited rising costs and labor shortages as reasons the industry must become more efficient and productive.
Journify raises $12M Series A amid 5x growthAd-signal infrastructure startup Journify raised a $12 million Series A at four times its previous valuation, led by Al Mada Ventures with Silicon Badia co-leading. The funds will support US and Europe expansion and platform AI; Journify serves more than 100 businesses.
White paper: embodied AI deployment hinges on system integrationAvnet, China's National and Local Co-Built Humanoid Robotics Innovation Center and GYH.AI released a 2026 white paper on embodied AI, identifying system integration, application-specific technology choices and lifecycle readiness as key to industrial deployment. An Avnet executive said progress now depends on how well technologies work together.
Researchers seek Chinchilla scaling experiment contactsA user asks whether researchers have run small-scale Chinchilla scaling-law experiments in PyTorch or JAX and invites discussion, linking DeepMind's paper. The post is a community query with no results reported.
Nexus Mods to require AI disclosure for new mod uploadsPC gaming mod site Nexus Mods will require new mod uploads to declare whether AI was used, addressing community concern that AI use is not clearly disclosed. Owner Victor Folmann said existing tags are insufficient and information must be clearer and easier to find.
DevRev delivers autonomous self-service with GPT-6.1 SolDevRev uses GPT-6.1 Sol to power Computer and Operator, balancing reasoning, speed and cost to deliver autonomous self-service for enterprise service.
Wall Street split on when the AI bubble will burstAt the 2026 Global Financial Leaders' Investment Summit in Hong Kong, Wall Street CEOs and big money managers disagreed on AI risks, including when the AI bubble might burst. Pimco CEO Emmanuel Roman was among those weighing in, but no consensus emerged.
CrowdStrike launches AI kill switch for rogue agentsCrowdStrike unveiled an AI Kill Switch at its Fal.Con Europe conference that can terminate misbehaving AI agents, block restarts and cut network access, available now in its Falcon Guardian platform. CEO George Kurtz called AI agent safety a major growth opportunity and cited an expanded partnership with Anthropic.
llama.cpp Community Shares Performance Data for Specific Quant and Hardware ComboA LocalLLaMA user posted llama.cpp fork results for prompt-processing and token-generation speeds under a specific quantization scheme and hardware setup, claiming strong but hard-to-reproduce numbers. The post lacks verifiable full configuration details and is a community anecdote of limited reference value.
Anthropic unveils FlashProve to verify AI model identityAnthropic is developing FlashProve, an open-source system that uses cryptographic proofs to verify an AI output came from the claimed model. Labs publish model fingerprints without revealing weights, preventing silent model swaps or false third-party claims.
Canva to cut and reshape about 22% of workforceCanva announced layoffs and an overhaul affecting about 1,271 people, or 22% of staff, who will lose jobs, change roles or relocate to central hubs. CEO Melanie Perkins framed the move as an AI-driven structural reckoning at one of Australia's highest-profile tech companies.
Data centers turn to retired jet engines as grid wait growsFacing years-long waits for grid connections and new turbines, some AI data centers are generating power with converted jet engines burning natural gas. xAI installed eight such turbines in Mississippi for Colossus 2, and OpenAI used gas turbines at its Texas Stargate site.
The A.I. Daily