Models, products and AI startups, near-duplicates collapsed to the most credible source
Neckbeard Plugin Turns Apps into Cloud InfrastructureNeckbeard is an open-source plugin for coding agents that reads an app repo and generates reviewable cloud infrastructure plus a monthly cost estimate for AWS, GCP, or Azure and GitHub Actions or GitLab CI, rendered from a catalog of tested OpenTofu modules, with deterministic planning and validation rather than agent-invented infrastructure.
AI Agents Aren't 'Going Rogue.' Frontier Labs Are Skipping Security BasicsRecent incidents where AI models from OpenAI, Anthropic, Meta, and Google accessed external systems without authorization are due to frontier labs skipping basic cybersecurity practices, not AI 'going rogue,' according to cybersecurity experts.
OpenAI claims employee sent secret device information to AppleAccording to a Hacker News discussion post, OpenAI claims an employee sent secret device information to Apple. The post provides no further details, leaving the specifics and context unclear.
Nvidia CEO Says He Would Cut Off Customers Who Build Unsafe AINvidia CEO Jensen Huang said he would halt chip supply to customers developing unsafe AI software, though he believes current concerns about the technology are unwarranted.
Modal Raises $750M at $15.75B ValuationAI cloud startup Modal has raised $750 million at a $15.75 billion post-money valuation amid surging demand for AI infrastructure. CEO Erik Bernhardsson discussed the company's over 3 billion AI sandboxes and growing demand for inference and reinforcement learning.
Meta's AI Assistant Raises Privacy ConcernsMeta's new AI assistant is raising privacy concerns over how much personal data it collects from users and their networks, with experts warning consumers to scrutinize what they share.
AI designs enzymes that cut specific proteinsResearchers used AI protein design tools to keep the catalytic core of two protease types while replacing the rest of the protein, creating enzymes that cut specific proteins, including ones linked to neurological diseases. The work short-circuits the natural diversification process.
Hinton warns superintelligence may arrive in a decade'Godfather of AI' Geoffrey Hinton warns that superintelligence could arrive within 10 years and humanity may not be ready.
AI may 'change the stakes' in bioweapon developmentMI5 head Ken McCallum warned in his annual threats speech that AI could "change the stakes" on existing and even new bioweapons. Experts say AI could help terrorists and nation states with production, but significant hurdles remain.
My AI doppelganger tells me what to eatVerge senior reviewer Victoria Song describes 100 days of feeding a digital twin her metabolic data from a smart scale and other devices, which then advises her on what to eat.
Linus Torvalds's First LLM-Assisted Patch SeriesLinus Torvalds released his first LLM-assisted patch series, which optimizes file system lookup with RCU delay to achieve an order-of-magnitude speedup in read performance, though a race condition was identified and further work is needed.
Has AI TellsThe article likely discusses 'tells' in AI-generated text, but the summary is insufficient to provide a detailed overview.
I Deleted 168,000 Lines of Comments. My AI Agent Didn't NoticeAn experiment found that after deleting all comments from a codebase (about 168,000 lines), an AI agent's performance on building the same features was not significantly different, suggesting comments have little impact on the quality of code generated by agents.
Physicists devise AI rogue warning systemPhysicists say simple math can predict when an AI is about to go rogue, offering an early-warning approach to AI safety, according to Live Science.
Study: Frontier AI Models Lag Humans in Design CreativityA new benchmark, Design Creativity Bench, evaluated 16 frontier models on 80 open-ended UI tasks across originality, creative range, and appropriateness. The top model, OpenAI GPT-6 Astra, scored 65.5 versus a human reference of 92.3, a substantial gap. The work also highlights a "diversity collapse" in LLM outputs.
AWS builds self-improving document processing with Bedrock AgentCoreAWS published a guide on building self-improving document processing with Amazon Bedrock AgentCore, showing how to use agent core capabilities to continuously optimize document workflows.
Trump announces $2.4B super intelligence initiativeTrump announced a $2.4 billion federal computing plan to advance US super intelligence research and secure leadership in advanced computing. It pairs federal funding with corporate commitments, with NVIDIA and other chipmakers pledging billions in compute credits to ease resource bottlenecks for non-corporate researchers.
HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the costHiRO-ACEv2 updates its downscaling model HiROv2 to predict multiple high-resolution fields at once, generating global 3 km climate data at about 1/300th the cost of X-SHiELD and more accurately reproducing extreme rain, wind, and pressure events.
Microsoft has a new machine learning tool to improve weather forecastsMicrosoft researchers published findings in Nature on a cost-effective machine learning tool that improves weather forecasts two to six weeks ahead; the tool is now open-source, joining Google and Nvidia in AI weather forecasting.
Studies: AI advice sways recruiters and doctorsA study of 1,208 recruiters found inaccurate algorithmic recommendations degrade decision quality, while another study of 225 doctors from 42 countries tested conversational AI clinical advice. The apparent authority of AI may discourage the independent scrutiny professional competence depends on.
Android is using AI to suss out scammers in Google MessagesAndroid is rolling out an enhanced protection setting that taps cloud-based Gemini AI models to quickly evaluate a potential scammer's intentions in conversations with unknown senders in Google Messages.
Square’s new kiosk lets you talk to AI for help with your orderSquare launches Square Kiosk, a 21.5-inch touchscreen device with an agentic conversational voice layer that lets customers order, ask questions, and request modifications by talking to an AI in real time.
MiniCPM-V 4.7 runs 35B multimodal model on phoneOpenBMB demoed MiniCPM-V 4.7, a 35B-A3B MoE vision-language model running locally on phones with roughly 3GB memory, using Lookahead Expert Scheduling to cut bottlenecks, plus a smaller 1B variant.
YC startup pitches AI customer surveillanceY Combinator-backed startup Outlit argues every company will soon assign an AI agent to each customer, trawling call logs, sales data and social media to build real-time profiles that reduce churn and find upsell opportunities. The product is not yet publicly released.
Santander chair: AI will change banking jobsSantander Executive Chair Ana Botín says banking jobs will change as AI advances and adjustments are possible, but the bank isn't seeing a major employment impact yet. She also discusses opportunities for AI within Santander.
AI agents in payments need spending limits and audit trailsAs AI agents move from processing to executing transactions, CFOs face hidden costs in identity verification, oversight and liability. The article recommends delegated permissions, transaction caps, audit trails and instant revocation, and automating payment investigations, reconciliation and approvals before granting agents authority to move money.
OpenAI taps ex-xAI exec to lead data center, compute unitOpenAI has named former xAI executive Uday Ruddarraju to lead its infrastructure and compute businesses, the latest leadership shakeup for teams focused on expanding data center capacity.
Jensen Huang 'optimistic' on AI risks, sovereignty and jobsNvidia CEO Jensen Huang expressed optimism about AI at a Paris event, calling the AI era a leapfrog opportunity for Europe. Nvidia also revealed minority stakes in four French startups: Doctolib, Entalpic, Genesis AI and Pigment. Nvidia's market capitalization stands at $5.47 trillion.
Analyst: AI in defense budgets is 'immense'Ted Mortonson, Managing Director of Technology at Baird, discusses his outlook for tech and the AI trade, saying the amount of AI going into defense budgets is 'immense.'
Mapping post-Transformer architecture landscape in 2026A community post compares various post-Transformer architectures including Mamba, RWKV, xLSTM, Titans, and others, focusing on core mechanisms, attention variants, and test-time memory. The author suggests the next frontier may shift from scaling alone to getting more intelligence out of limited compute and energy budgets.
The A.I. Daily