Models, products and AI startups, near-duplicates collapsed to the most credible source
Walmart launches Agent Gateway for AI agent shoppingWalmart launched Agent Gateway, an MCP-based server from Anthropic's protocol, letting approved AI agents search inventory and buy for users with spending and transaction controls across Walmart and Sam's Club. It contrasts with Amazon's blocking of AI agents; Walmart already partners with Meta's Muse and integrates ChatGPT and Gemini.
Mistral releases open-weight multimodal model Large 4Mistral AI released Mistral Large 4 in public preview, an open-weight, general-purpose multimodal model with a fine-grained Mixture-of-Experts architecture: 52B active parameters, 1.05T total parameters, and a 1.6B vision encoder.
Harvard mathematicians split over OpenAI math breakthroughAfter OpenAI said it solved the roughly 90-year-old Navier-Stokes problem, Harvard mathematicians disagreed: some worry AI proofs lack citations and explanations and are hard to evaluate, while others say accuracy matters more than origin. A September-formed advisory group on math and AI aims to guide how AI research is released.
Google staff resign over military AI contractsBBC spoke to several Google DeepMind staff who resigned or are leaving over the company's AI contracts with the US and Israeli militaries; London research engineer Sofia Liguori resigned this week. Google said contracts go through compliance reviews and that it has channels for employee concerns.
Anthropic launches proactive agent Claude DuetAnthropic launched Claude Duet, a proactive agent powered by Opus 5.5 that keeps working while users are away and waits for approval when needed. Available to Pro and Max plans in the US, it works with more than 6,000 connectors and plugins.
AI Will Become A Commodity Over Time: NamburiMadhu Namburi, managing director at General Catalyst, said AI will become a commodity over time. He discussed AI's impact on dealmaking and investments on Bloomberg.
Anthropic publishes public benefit charter with five pillarsAnthropic, a Public Benefit Corporation, published a charter and report committing to balance shareholder interests with public benefit. It outlines five pillars: solving AI safety, universal safety practices backed by law, democracy and global stability, prosperity and flourishing, and shared benefits.
NVIDIA launches Kumo Relational foundation modelNVIDIA Kumo Relational is a pre-trained relational foundation model that predicts directly on enterprise databases without task-specific training or feature engineering. Across 41 tasks on 15 databases it beats supervised relational deep learning and expert-tuned gradient-boosted trees, and on SAP SALT it tops AutoGluon ensembles by 8% in mean reciprocal rank. It scales to 500B+ rows and is available via an open-source library or NVIDIA NIM.
Judge rejects Trump claim EPA can dismiss data center lawsuitsA US judge rejected Trump's claim that the EPA can unilaterally dismiss Clean Air Act lawsuits, so xAI (now SpaceXAI) still faces a suit over alleged pollution from its Memphis Colossus data center. The NAACP seeks best-available pollution controls and penalties up to $124,426 per day per violation.
Scentronic: AI-assisted interface for remote odor sharingScience published research on Scentronic, an AI-assisted olfactory interface for remote odor sharing and transmission, exploring AI's role in digitizing and remotely delivering smells.
User runs 748B DeepSeek-V4.1-Flash on a €2k PCA user ran the 748B-parameter DeepSeek-V4.1-Flash-1017 at 7.2 tok/s on a used ~€2,000 PC (RTX 3090, 128GB DDR4, i9-9900X) using a 1-bit quant and llama.cpp. A 110GB Engram table stays on NVMe and is read on demand, letting it fit. The output contained a physics error and has little practical use, the user said; Qwen models work better on the setup.
G.M. Hopes to Design New Cars in as Little as 2 Years Using A.I.General Motors aims to use artificial intelligence to cut new car design time to as little as two years. The move could significantly speed up vehicle development and reduce costs, though details on implementation and investment were not disclosed.
Major AI firms sharply increase lobbying spendingWatchdog group POGO reported that major AI companies are sharply increasing their lobbying spending, reflecting the industry's growing push to influence policymaking.
Train Microduck to walk in Isaac Lab 3.0 and deploy to real robotPollen Robotics and Hugging Face released Microduck, an affordable open-source bipedal robot, with a tutorial for training walking policies in GPU-parallel simulation using NVIDIA Isaac Lab and Newton. Actuator modeling and domain randomization prepare policies for real hardware; policies export to ONNX for deployment, can be shared on the Hugging Face Hub, and can compete in Microduck Arena.
Researcher Proposes Noun-Verb-Conjunction Triad Internal LanguageA community post proposes giving neural networks a structured 'language of thought' built from noun-verb-conjunction triads, with each attention head holding one triad and understanding built up in sentence-sized units rather than one giant graph. The idea is conceptual and untested.
Florida lawyer referred for discipline over 'AI slop' filingA Florida lawyer was referred for possible discipline after filing court documents that judges called 'AI slop.' The case highlights growing judicial scrutiny of low-quality or inaccurate AI-generated submissions.
Websites scramble to fix software flaw that could put much of internet at riskWebsites are rushing to patch a software flaw that could put much of the internet at risk. The vulnerability, if exploited, could cause widespread security problems. Details of the flaw and affected software were not disclosed.
Data center boom sparks debate over benefits, costs and risksU.S. data center investment is surging, boosting the economy while drawing opposition. Global capacity tripled from 2015 to 2024 and accelerated after ChatGPT's launch. Supporters call data centers the infrastructure backbone for AI; critics warn of a debt-financed bust, exaggerated local jobs and tax benefits, energy and environmental costs borne by communities, and risks from AI itself.
Specializing SLM vocabularies can cut inference cost by 77%A preview of the VocabX paper shows that rebuilding the Qwen3-4B vocabulary for a single domain and output format can greatly cut inference cost. In a Portuguese essay-grading system, the method designed the vocabulary around token cost rather than frequency; with the same base model, data and training recipe, measured inference cost fell 77%, while shrinking the roughly 151,669-entry vocabulary scored at each generated token.
Blacksmith raises $66M Series C at $1B valuationBlacksmith, a platform for building and testing AI-written code, raised a $66M Series C led by Sequoia at a $1B valuation. ARR grew 17x and it has run over 90 million CI jobs; it was valued at $550M just two months ago.
Report flags legal risks of AI-generated drug evidenceA Crowell & Moring client alert says probabilistic AI models in drug discovery can produce plausible-but-wrong outputs that affect regulatory submissions, trial integrity and legal liability. It lists five risks: hallucinations, unrepresentative training data, opaque reasoning, input sensitivity and adversarial manipulation of training data, and urges experimental validation, human oversight and preparation for FDA and EU guidance.
Poll: 57% of Americans say AI is bad for societyAn ABC/Washington Post/Ipsos poll finds 57% of Americans say AI is bad for society versus 42% good, and about two-thirds oppose new US data centers, with stronger opposition locally. Among those under 30, 65% view AI negatively, while Republicans are more evenly split.
Hassabis shortens AGI timeline to two to three yearsDeepMind founder and Nobel laureate Demis Hassabis now expects artificial general intelligence within two to three years, down from three to five years earlier this year, marking another pull-forward of his forecast.
AI Research Startup Flapping Airplanes Raises $1B at $6B ValuationOne-year-old AI research startup Flapping Airplanes raised $1 billion at a $6 billion valuation, led by Index Ventures and Lux Capital, with Sequoia, Alphabet's GV and Kleiner Perkins participating. The company has yet to launch a commercial product, and its valuation has quadrupled.
AT&T rethinks in-house legal work with OpenAIAT&T is working with OpenAI to make legal knowledge easier to access. Its LegalEdge team is rethinking legal intake and scaling legal judgment across the organization.
Omnara Releases Open-Source Durable Agent APIOmnara has launched an open-source API for running durable AI agents, with sandboxes, secrets management, permissions, and RBAC, storing state in Postgres so agents survive restarts and cost nothing when idle. Positioned as an open-source alternative to Claude Managed Agents and the OpenAI Agents API, it can be self-hosted with users' own models.
Rust off Rails: a web stack written entirely by AI agentsRust off Rails is an experimental web stack consisting of Markdown specs that tell coding agents how to build a full-stack Rust app; agents write all code while humans supply only intent and specs. The project drew HN discussion.
Latin America and Caribbean urged to take ownership of AIECLAC and Chile's CENIA released the fourth Latin American AI Index (ILIA 2026), analyzing 20 countries and, for the first time, comparing Germany, Spain, Estonia, Portugal and Singapore. The report says the region's biggest gaps are computing capacity, business AI adoption and national strategy implementation; connectivity alone is insufficient, and without linking infrastructure, data, talent, research and institutions, dependencies and inequality may deepen.
China Reportedly Doubling RAM Output to Curb Gadget Price SurgesA Reddit discussion links to a report that China is doubling RAM production to keep everyday gadget prices from skyrocketing. The post offers no specific figures or timeline.
GitHub migrates storage to Azure for agent-scale Git trafficGitHub said agentic AI drove exponential Git traffic growth; nearly all repositories now live in both its data centers and Azure. Constrained by local power capacity, GitHub moved storage to Azure, which now serves about half of Git traffic, and changed quorum write serialization.
The A.I. Daily