The A.I. Daily
LiquidAI Releases d1-3B Model on Hugging Face
LiquidAI released the d1-3B model on Hugging Face; a community user shared the model page, with capability and benchmark details not yet disclosed.
Industry
Everything · newest firstModels, products and AI startups, near-duplicates collapsed to the most credible source
Agent.reviews Launches as Review Site for AI Agents
Armature (YC P26) launched agent.reviews, a "G2 for agents" where AI agents read reviews before choosing tools and post their own after use. Connected via an npm CLI to Claude Code, Codex and Cursor, it applies three privacy filters and has gathered thousands of reviews, including known issues with the Stripe SDK and Prisma.
Liquid AI releases open edge decision models
Liquid AI released two open decision models: d1-3B and experimental d1-omni-600M. d1-3B scores 48.57 on Decision Index 0.2.1, ahead of all 4B and 9B models, and answers in 16 ms on NVIDIA Jetson AGX Thor; it handles text and images, while the omni model also supports audio.
San Francisco passes data center moratorium
San Francisco, home to some of the world's most valuable AI companies, has temporarily banned new data centers within its borders. The report did not detail the ban's duration or scope.
Anthropic Shifts to R&D, Eyes 100K sf Mission District Space
Anthropic is shifting its real estate focus from offices to R&D and is eyeing roughly 100,000 square feet of industrial space in San Francisco's Mission District, according to The Real Deal, signaling heavier R&D investment.
Meta rolls out AI tools to detect child abuse ads
Meta said it took action against 33.2M pieces of child sexual exploitation content on Facebook and Instagram in H1 2026, over 97% detected by its systems before user reports. New tools include an LLM system to detect 'signposting' ads and a red-teaming AI agent testing its own safeguards.
Anthropic dispute highlights gaps in Pentagon AI oversight
Congress has not enacted legislation specifically governing the DOD's use of frontier AI models or their reliability in autonomous weapons. Amid the Pentagon's dispute with Anthropic, President Trump in February directed federal agencies to stop using its technology and Defense Secretary Hegseth ordered it excluded from the supply chain; on Sept. 25 a D.C. Circuit panel upheld the exclusion. Anthropic CEO Amodei said frontier AI is not reliable enough for fully autonomous weapons.
Marvell Investor Day: Connectivity lifts outlook to $20 billion
At its investor day, Marvell raised its outlook, with connectivity as the main driver lifting its overall target to $20 billion. No further financial details were disclosed.
Joe Tsai: Open Source Is Europe's Best Path to AI Independence
Alibaba Chairman Joe Tsai said open-source AI is Europe's best route to independence from US and Chinese technology. He urged the region to build its own computing infrastructure to train and run AI models.
Five ways to make the new US AI task force succeed
The White House launched the Super Intelligence Force to implement the White House Accord on Super Intelligence, under which Trump and the heads of Anthropic, Google, Meta, Nvidia, OpenAI and SpaceXAI committed to internal security controls, independent safety evaluations and board oversight. A former NSA chief AI officer offers five recommendations.
National Compute CEO on building a shared AI compute grid
National Compute CEO Anjney Midha discussed on Bloomberg his plan to turn AI infrastructure into a shared national grid, opening powerful chips to students, researchers, and startups nationwide.
Meta's Muse app launches on iPad
Nearly a month after its iOS debut, Meta's agentic AI tool Muse adds native iPad support, following a Mac version. The app spent weeks as the top free app on Apple's App Store; the iPad version better uses the larger screen and multitasking.
OpenAI launches Decisions API for simple classification
OpenAI launched a Decisions API that classifies text and images into yes/no probabilities, category picks, or scale ratings, about ten times faster than the Responses API at $0.10 per million input tokens, and cut paid API tiers from five to three.
Pentagon speeds up 'kill chain' AI buys with 5-minute videos
The Pentagon's Tradewinds initiative is making it easier to award millions to "nontraditional" defense contractors including OpenAI, Anthropic, and Google, using 5-minute videos to speed up procurement of AI for the "kill chain."
India AI Rich List: 19 People Hold Rs 3.07 Lakh Crore
The M3M Hurun India Rich List shows 19 AI-sector individuals with combined personal wealth of Rs 3.07 lakh crore, 15 of them new this year. CTRL S's Sridhar Pinnapureddy leads with Rs 46,100 crore, while former Twitter CEO Parag Agrawal enters via Parallel Web Systems at Rs 10,500 crore.
HUMXN offers free home services in Minnesota to train robots
AI data firm HUMXN is partnering with home-service companies in Minneapolis, Chicago and Miami to offer qualifying homeowners free plumbing, electrical and HVAC services while turning technicians' work into training data for AI and robotics.
Healthleap raises $38M for AI that flags at-risk hospital patients
Healthleap raised $38M across seed and Series A rounds backed by Sequoia, First Round, and Hummingbird Ventures. Its AI plugs into hospital EHRs to flag undiagnosed conditions like malnutrition and delirium from clinician notes, and is deployed in over 50 hospitals.
Meta's Muse hits 6.6M downloads, 1.8M daily users
Meta's Muse has topped 6.6 million downloads and 1.8 million daily active users since its Sept. 8 launch, showing consumers will trade personal information for AI assistance. Sensor Tower's Seema Shah discussed the growth of personal AI agents.
Only 4% of US CHROs feel very confident leading AI change
Only 4% of US CHROs said they felt very prepared to lead AI-driven workplace change over the next two years, per The Conference Board's quarterly index. Most firms changed workforce planning due to AI, but just 19% folded AI's expected effects into enterprise-wide workforce and financial planning, citing difficulty predicting impacts and limited AI skills.
FCC considers allowing more political robocalls before midterms, including AI-created ones
The FCC is considering allowing more political robocalls before the midterm elections, potentially including those created with AI. The report offers no specific rule details or timeline.
Tony Fadell: first AI gadgets failed to solve real problems
At MIT Future Fest, iPod creator Tony Fadell said first-gen AI gadgets like the Rabbit R1 and Humane Ai Pin failed because they solved no real pain point. He noted most consumers have never used an assistant and that the next wave must earn trust by addressing actual needs.
Google, Meta, Isomorphic invest $300M in virtual cell research
Google DeepMind, Meta, and Isomorphic Labs are jointly investing $300 million in Biohub, the nonprofit founded by Mark Zuckerberg and Priscilla Chan. Part of a $1.8 billion initiative, the funds aim to build AI datasets and a "virtual cell" for simulating disease prevention and management.
SpaceX vs. Nvidia: The Better AI Stock for 2027?
The Motley Fool compares SpaceX and Nvidia, arguing the key is which can more predictably turn AI spending into profit. Nvidia's Q2 revenue hit $96.2 billion, up 106% year over year, with data center at $89 billion, as it transitions from Blackwell to the Vera Rubin platform.
OpenAI's Alexander Embiricos to speak at TechCrunch Disrupt 2026
Alexander Embiricos, OpenAI's product lead for Dots, will speak on the AI Stage at TechCrunch Disrupt 2026 about what comes after the chatbot. Dots connect to everyday apps and work toward goals in the background, raising key questions about autonomy, privacy, and trust.
User complains Apple AI models can't be removed on macOS 27
A Hacker News user said macOS 27 still occupies about 30GB of local AI models even after disabling Siri and Apple Intelligence, cannot move them to iCloud, and called for more negative coverage to get Apple's attention.
A Week in Beijing and Shanghai With China's AI Builders
An article about the people building AI in Beijing and Shanghai was shared in the LocalLLaMA community for interested readers.
Google launches SynthID Detector
Google launched SynthID Detector, a tool to identify AI-generated content such as images, audio and text carrying SynthID watermarks, aiming to improve transparency and provenance of AI content. The blog post disclosed few further technical or availability details.
AI-assisted proof of optimal packing for 11 squares
An AI-assisted effort produced a complete optimality proof for packing 11 squares, with all 7,920 local Lean modules accepted and a final audit showing zero admissions, while geometry and checker soundness retain ordinary Lean proofs.
Penn Professor: 0.48% Chance AI Kills a Billion by 2030
A University of Pennsylvania professor estimates a 0.48% chance that AI causes the death of a tenth of humanity by 2030. The figure is a personal estimate; the report offers no further methodology or basis.
TechCrunch Disrupt 2026 Opens in Six Days
TechCrunch Disrupt 2026 will convene more than 10,000 attendees at San Francisco's Moscone West in six days. Registering before doors open saves up to $100 per pass and gets 50% off a second pass of the same type, as on-site prices rise.