Today's AI Pulse
Anthropic shuts down Claude Fable 5 and Mythos 5 after U.S. export‑control warning
Anthropic disabled access to its Claude Fable 5 and Claude Mythos 5 models after the U.S. Commerce Department warned that foreign nationals could exploit the systems. The warning followed alerts from Amazon and other tech firms about a potential jailbreak that could facilitate cyber‑security attacks. Anthropic declined the administration’s request for a voluntary shutdown.
Also developing:
By the numbers: Meta scraps $2B Manus acquisition amid China divestiture order
Google Overhauls Search With 24/7 AI Agents: What It Means for Affiliate Traffic
Google unveiled a major overhaul of its Search box at I/O 2026, converting it into an AI‑driven workspace that accepts long, multimodal queries and integrates personal data from Gmail and Photos. AI Mode now serves over one billion monthly users, with usage doubling each quarter, while new 24/7 information agents for AI Pro and Ultra subscribers continuously monitor web content and deliver task‑focused updates. The changes push search toward zero‑click experiences, threatening thin affiliate pages but rewarding publishers with original data, structured markup, and verifiable expertise.
OpenAI’s Reasoning Model Disproves 80‑Year‑Old Erdős Conjecture
OpenAI announced that its latest general‑purpose reasoning model produced an original proof that disproves a famous geometry conjecture first posed by Paul Erdős in 1946. The breakthrough, described as the first time AI has autonomously solved a prominent open problem,...
The Sequence Opinion #864: Every AI Agent Needs a Computer
The post argues that the next breakthrough for AI agents lies not in larger models or longer context windows, but in giving each agent direct access to a computer. An agent confined to token output is likened to a "brain...
AEON Announces $8M in Funding Led by YZi Labs to Build the Settlement Layer for Agentic Economy
AEON announced the close of an $8 million pre‑seed round led by YZi Labs, with participation from IDG Capital, HashKey Capital and other investors. The startup has already launched an AI‑payment product that connects autonomous agents to more than 50 million real‑world...
Revive Unveils Next‑Gen AI Valuation Platform as Rayse Adds Real‑Time MLS Tools
Revive rolled out a major upgrade to its AI‑driven home‑value platform, adding Smart Value Comparison, room‑by‑room condition analysis and renovation modeling. At the same time, client‑communication startup Rayse announced a partnership with Momentum MLS to embed its AI assistant and...
State AGs Urged to Crack Down on AI 'Nudify' Apps in Bipartisan Letter
A bipartisan coalition of 54 advocacy groups, led by UltraViolet, sent a letter to state attorneys general demanding legal action against app stores that host AI-powered nudification apps. The letter cites 55 apps on Apple’s store and 47 on Google...
NanoCo Raises $12M Seed as It Launches Docker‑Sandboxed AI Agents for Every Employee
NanoCo announced a managed enterprise AI service that provisions a personal, Docker‑sandboxed agent for every employee, while closing a $12 million seed round led by Valley Capital Partners. The launch targets large firms seeking secure, per‑user AI assistants, differentiating itself from...
Utah Residents Challenge $100 B AI Data Center Over 16 Billion‑Gallon Water Demand
Residents and environmental groups in Box Elder County, Utah, rallied against the Stratos Project after county commissioners approved the $100 billion, 9‑gigawatt AI hyperscale data center on May 4, 2026. Opponents warn the campus could consume 4.24‑16.6 billion gallons of water per year,...
8 AI Costs Leaders Don't Always Budget for but Should
Enterprises often focus on AI infrastructure, licensing and cloud costs, but eight hidden expenses can dwarf those line‑item budgets. Failed pilots, endless experimentation, retrofitted governance, talent turnover, human‑in‑the‑loop oversight, model drift, inference lock‑in, and reputational risk each add sizable, hard‑to‑measure...
Stony Brook AI Vision System Boosts Recycling Sorting Accuracy
Stony Brook University researchers have built an AI‑driven vision system that can identify paper, plastics, food waste and fabrics in municipal solid waste, delivering faster, higher‑accuracy sorting than manual methods. The project, backed by a university AI Innovation Seed Grant,...
Citrix Warns of Rogue AI as Enterprise Attack Surface Grows
Citrix warns that rogue AI—autonomous systems operating outside governance—poses a growing threat as enterprises expand automation and digital workspaces. The company says AI agents can bypass traditional security perimeters, increasing the attack surface and enabling faster, more sophisticated cyber‑attacks. Citrix...
Modern AI Is Often Judged to Be More Human than Actual Humans in Turing Test Experiments
Researchers Jones and Bergen published a study in PNAS showing that modern large language models can pass a classic three‑party Turing test. When given a detailed persona prompt, GPT‑4.5 was judged human 73% of the time and LLaMa‑3.1‑405B 56%, far...
The Future of Physical AI Isn’t Smarter Robots, It’s Smarter Interfaces
Physical AI has focused on smarter robots, but Wetour Robotics argues the next breakthrough lies in smarter human‑machine interfaces. The company’s Orchestra platform uses Spatial Intent Fusion to combine vision, spatial positioning, and sEMG biosignals, delivering intent commands in under...
Ace Hardware Debuts AI-Powered Assistant
Ace Hardware has introduced an AI‑powered assistant called “Hey ARMA” for store associates, accessible on handheld devices. The tool delivers instant product knowledge, project guidance, and recommendations, allowing staff to spend less time searching for answers. Hey ARMA is already active in...
AI Has Striking Science Skills, but Grad Students Are Still Wanted
Physicist Scott Dodelson reports that agentic AI systems like OpenAI Codex can perform graduate‑student tasks—literature review, coding, data fitting, and plotting—in seconds, delivering results that would take months for a human. He cites a Harvard colleague who claimed a ten‑fold...
Has AI Conquered Coding? (It’s Not So Simple…) | AI Reality Check
In this episode Cal Newport examines the hype around AI‑driven "agentic coding" by dissecting a recent essay from programmer Lars Faye, who warns that relying on AI to write code can erode developers' critical thinking and skill. Newport highlights real‑world...
How Can We Prevent AI Models From Cannibalizing Themselves when Human-Generated Data Runs Out? Scientists Say They've Found the Answer.
Scientists warn that large language models may soon exhaust high‑quality human‑generated data, risking a phenomenon called model collapse where AI outputs become nonsensical. A study published in Physical Review Letters shows that inserting a single human‑crafted data point into a...
MIA Unveils AI-Powered Airport Operations Center and Digital Monitoring Hub
Miami International Airport will complete a $33 million, 13,254‑square‑foot AI‑enabled Airport Operations Center and Digital Monitoring Hub by 2027, the first of its kind in the U.S. The facility features AI‑powered long‑range PTZ cameras, real‑time digital tower technology and a 360°...
Machine Learning System Design Interview #33 - The Streaming Bias Trap
In a Meta senior ML‑Ops interview, candidates are asked to uniformly sample an unbounded, real‑time event stream into a fixed‑size buffer. The correct solution is reservoir sampling, which mathematically guarantees each event has equal probability of selection. Naïve approaches like...
Free, Self‑hosted AI Search Replaces $20 Perplexity Subscription
A team in San Francisco killed Perplexity's $20/month subscription. It's called Vane. You get AI-powered search with cited sources, follow-up questions, image and video search, and focus modes for academic papers, Reddit, YouTube, and Wolfram Alpha, running entirely on your own...
SPEAKER INTERVIEW: Mircea Gradu, VP of Vehicle Engineering, Karma/Chair of FISITA Board
Mircea Gradu highlighted the rapid move toward safety‑case‑based regulations for automated driving, noting UNECE's February 2026 draft and the U.S. DOT's new audit framework. He detailed a three‑phase validation pipeline—simulation, X‑in‑the‑loop testing, and live deployment at 25 Orange County intersections—that...
AI in Compliance Recording: Five Questions Firms Need to Answer Now
Capital‑markets firms now face a sprawling communications landscape that includes voice, chat, collaboration tools and mobile channels, forcing them to rethink compliance recording beyond simple retention. A‑Team’s whitepaper outlines five critical questions, from full‑channel capture and AI‑ready architecture to moving...
INTERVIEW: Matt Fisch, Chairman and CEO, AEye
AEye CEO Matt Fisch describes the company’s mission to give machines sight through advanced lidar, positioning it as a cornerstone of the emerging "physical AI" era. The firm’s ultra‑vision platform offers long‑range, high‑resolution perception that outperforms traditional cameras and radar,...
This Unexpected Body Part Could Reveal Early Signs Of Bone Loss
A new AI system called RetiAGE analyzes retinal photographs to estimate biological age and predict osteoporosis risk. The study, which examined 1,965 older adults in Singapore and 43,938 participants from the UK Biobank, found that eyes appearing older than a...
Microsoft Clarity Now Shows Grounding Queries Behind AI Citations via @Sejournal, @TaylorDanRW
Microsoft Clarity now displays the grounding queries that AI engines like Copilot use to retrieve and cite website content, giving SEOs a clear view of the exact search terms driving AI citations. The feature revealed a case study with 36,000...
AI at Scale: What Engineering Teams Are Confronting
Enterprise AI optimism has given way to hard‑won lessons about production readiness. While three‑quarters of IT leaders now train models and 76% run GPU workloads in live environments, they face friction integrating agentic AI into legacy cloud stacks. Governance gaps...
Scaling the Memory Wall: HBM, CXL, and the New GPU Playbook
The AI data‑center market is confronting a growing memory wall as inference workloads demand ultra‑low latency and massive KV‑cache capacity. High‑bandwidth memory (HBM) remains the primary bandwidth solution, but supply constraints and thermal costs are driving prices up, with the...
The Dark Data Problem Hiding Inside Your AI Agents
OpenClaw recently broke the GitHub record with 250,000 stars, prompting NVIDIA CEO Jensen Huang to label it the "operating system for personal AI." While NemoClaw adds kernel‑level sandboxing for runtime governance, it leaves agent‑generated data vulnerable to loss. Autonomous agents...
Amazon.com, Inc. (AMZN) Expands with AI-Powered Shopping Guides
On May 13 Amazon introduced Alexa for Shopping, an AI‑powered assistant embedded in its main search bar that can answer product queries, compare specifications, and build personalized buying guides. The tool works on both the Amazon website and app without...
Automating Knowledge Work
RAND’s May 2026 paper introduces a systematic framework for automating knowledge work, focusing on tasks that involve analysis, decision‑making, and information processing. It evaluates each task across four dimensions—criticality, accuracy, novelty, and observability—to determine the appropriate level of AI oversight. Based...
MaxMine Deploys Production-Grade Machine Learning System at Australian Mine Sites
MaxMine has rolled out a production‑grade machine‑learning system for load‑and‑dump classification across several Australian mining operators, including Glencore, NRW Holdings and Macmahon. The solution, now six months live, leverages more than 14 million hours of labelled operational data to deliver higher‑accuracy...
Claude for Legal and Access to Justice: The Good, the Bad, and the Unknown
Anthropic unveiled a major upgrade to its Claude AI, adding over 20 legal‑specific connectors and 12 practice‑area plugins that integrate directly with Microsoft Word, Outlook, Excel and PowerPoint. The rollout targets law firms, corporate legal departments, and legal‑tech providers seeking...
Specialization Emerges Naturally in MoE Training
MoE Training, Part 1 — in one tweet: You do NOT assign "this expert handles medicine, this one handles law." You start with 9 random experts + a router. The router learns to pick 2–3 per question. Specialization emerges from data, not...
An MCP Connection to Claude May Soon Be a Necessity for Legal Tech Companies and Their Customers
Trellis Law, an AI‑driven legal research platform focused on state trial courts, has launched an integration with Anthropic’s Claude using the Model Connectivity Protocol (MCP). The service aggregates trial‑court data from 45 U.S. states, structuring judges, opposing counsel, dockets and...
Who Is Andrej Karpathy? The AI Researcher Behind Tesla Autopilot, OpenAI and the Course that Taught Millions
Andrej Karpathy, a Stanford‑trained AI researcher, has been instrumental in shaping modern artificial intelligence through breakthroughs in image captioning, neural‑network interpretability, and generative models. He co‑founded OpenAI in 2015, helping set the research agenda that later produced the GPT series. At...
AI & Energy: Bending The Curve
Artificial intelligence is expanding at a pace that forces data‑center power consumption to rise 4–5 times each year, pushing installations toward megawatt and eventually gigawatt scales. This surge strains grid capacity and makes energy the primary bottleneck for AI performance. Industry...
Why Are Students Opening Up to AI Instead of People?
The Rithm Project surveyed nearly 2,400 U.S. teens and young adults and found that roughly one‑third use AI for emotional or relational support, often after an initial academic use. Students describe AI as a private, non‑judgmental outlet that fills acute...
Confidential Inference Now Available for Open and Closed Models
Confidential inference means you have end-to-end encryption with nobody else getting access to your prompts. But this only works on open weight models (for now). If you do still want ot use closed weight models - you can use it via...
KFintech Launches AI-Powered Investor Relations Platform AEGIX for Listed Companies
KFin Technologies launched AEGIX, India’s first AI‑native investor relations platform, targeting companies listed on the NSE and BSE as well as pre‑IPO firms. The solution combines ownership analysis, market intelligence, investor targeting and an AI assistant named AiRa into a...
Acer Veriton GN100 Packs NVIDIA Blackwell AI Compute Into Compact Workstation
Acer unveiled the Veriton GN100, a compact AI workstation built around NVIDIA’s DGX Spark platform and the GB10 Grace Blackwell Superchip. Despite measuring only 150 mm × 150 mm × 50.5 mm, the system delivers up to 1 petaFLOPS of AI performance, 128 GB of unified LPDDR5x memory, and a 4 TB...
What Banks Are Doing with AI
Banks are quietly embedding generative AI into middle‑office functions rather than replacing core systems. Anthropic, OpenAI, Google Gemini, Microsoft Copilot, IBM watsonx and AWS Bedrock each target specific workflow bottlenecks such as KYC, AML, document processing and operational automation. Real‑world...
Why Deepfake Fraud Beats Your Workflows, Not Your Technology - with Jon-Rav Shende of Thales Group
In this episode, Global CTO for Data and AI at Thales Group, Jean‑Rav Shende, explains that deep‑fake voice fraud exploits the human workflow of contact‑center agents rather than just the underlying technology. He highlights that the greatest risk occurs where...
Ten AI Cyber Vulnerability Questions - NCSC
The UK National Cyber Security Centre (NCSC) released a checklist of ten questions for organisations that deploy AI models to discover cyber‑security flaws. It stresses that AI‑driven detection is only valuable if basic security hygiene—such as patching and access control—is...
Lenskart to Step up AI Push in FY27, Targets Scale-Up in Eye Tests, Automation
Lenskart announced that artificial intelligence will drive its FY27 growth strategy, rebranding itself as a consumer‑AI company. The retailer plans to scale eye‑testing capacity to 100 million tests, up from roughly 30 million today, and embed AI across stores, factories, and its...
Alibaba Announces Comprehensive Full Stack AI Upgrade
Alibaba unveiled a full‑stack AI upgrade at its Cloud Summit, introducing the Qwen 3.7‑Max large language model designed for long‑running, multi‑tool agent tasks. The company also launched the Panjiu AL128 Supernode server, packing 128 AI accelerators and petabyte‑per‑second rack bandwidth. Its semiconductor...
Point72’s AI Infrastructure Play: Steve Cohen’s Firm Pushes Deeper Into the Operating System of Modern Finance:
Point72 is investing in Mercury, a fintech that raised $200 million at a $5.2 billion valuation and serves over 300,000 startups. The hedge fund views AI infrastructure as a core operating layer, extending beyond hardware to the financial systems that power AI‑native...
Google AI Edge Gallery: Local AI Agents Get Tools, Memories, and Memory
Google has upgraded its AI Edge Gallery with experimental Model Context Protocol (MCP) support, local schedule notifications, and persistent chat histories. MCP lets on‑device models decide which external tool to invoke, then routes the call to a local or cloud server while...
WebMCP: Google Wants to Make the Web Machine-Readable for AI Agents
Google unveiled WebMCP at I/O 2026, an open‑web proposal that lets developers expose JavaScript functions and HTML forms as structured tools for browser‑based AI agents. The experimental Origin Trial launches with Chrome 149, and Google’s Gemini model in Chrome is slated to...
How to Prompt ChatGPT: 10 Tips for Better Answers
The article outlines ten practical tips for prompting ChatGPT and similar AI chatbots, emphasizing a structured prompt skeleton that defines role, task, context, and output. It stresses the need for concise context, illustrative examples, and explicit length or format specifications...
How AI Is Shaping Biotech? Expert Insights From James Bilotta
In this episode of Biotech Bytes, SVP and Chief Digital Information Officer Jim Bellotta discusses how AI and digital transformation are reshaping biotech, emphasizing the shift of the CIO role from a back‑office tech function to a strategic business partner...