Today's AI Pulse
-1.png)
Anthropic shuts down Fable 5 and Mythos 5 under U.S. export controls
Anthropic abruptly disabled its flagship Fable 5 and Mythos 5 models to comply with a U.S. export‑control directive, exposing the vulnerability of firms that depend on closed‑source AI services. The move sparked a sharp rally in Chinese open‑source offerings such as MiniMax and Zhipu as investors shift to downloadable, on‑premise models.
Also developing:
By the numbers: CNTXT AI raises $60M Series A
Rice and MD Anderson Unveil PrecisionView Handheld AI Microscope for Early Cancer Detection
Scientists from Rice University and MD Anderson Cancer Center introduced PrecisionView, a handheld AI‑powered endomicroscope that delivers cellular‑level images across a field of view five times larger and a depth of field eight times greater than conventional devices, aiming to accelerate early cancer detection at the point of care.
MethodHub Unveils QuantumHire, AI‑Powered Platform Targeting Enterprise Recruiters
MethodHub launched QuantumHire, an AI‑driven full‑time hiring platform designed for enterprise recruiters. The solution promises to automate screening, improve candidate experience, and reduce reliance on spreadsheets and disconnected systems, signaling a new competitive push in the HRTech market.
Dataiku Adds AI Builder to Snowflake, Bedrock Data Offers Free Cortex Governance
Dataiku announced Cobuild on Snowflake, a governed AI workflow builder that integrates Snowflake Cortex services with Dataiku’s orchestration platform. At the same time, Bedrock Data introduced a no‑cost AI governance solution for Snowflake Cortex, giving enterprises additional oversight tools.
Dell Launches Slim AI‑Powered 14S and 16S Laptops in Copilot+ Lineup
Dell introduced the 14S and 16S thin‑and‑light laptops, expanding its Copilot+ AI PC family. Both models run Intel Core Ultra 9 386H chips, deliver up to 50 TOPS of on‑device AI performance, and promise up to 26 hours of streaming battery...
Point72’s Turion AI Fund Posts 15% April Gain, Defining AI Infrastructure Trade
Point72’s AI‑focused Turion fund, managed by Eric Sanchez, posted a 15% gain in April, outpacing the broader technology hedge‑fund category. The performance underscores the shift toward AI infrastructure assets—semiconductors, data‑center hardware, and networking—as the primary source of hedge‑fund alpha.
OneStream and Microsoft Deepen AI Partnership to Empower CFOs
OneStream and Microsoft have expanded their strategic partnership, pledging a multi‑year investment to scale AI infrastructure for the Office of the CFO. The collaboration will deepen integration of OneStream's SensibleAI tools into Microsoft 365, aiming to accelerate forecasting, anomaly detection...
Y Combinator and Z Fellows Back 19‑Year‑Old Delhi AI Founder
Y Combinator and Z Fellows have admitted 19‑year‑old Samrath Singh Chadha, a Delhi‑based AI developer, into their accelerator programs. The teen’s journey from a failing computer‑science grade to building large‑language‑model tools earned him $1,000 in Google cloud credits and a...
BoomiWorld 2026 Unveils AI Data Trust Suite and Governance Tools
At BoomiWorld 2026, Boomi rolled out a suite of AI‑focused data activation products—including Boomi Connect, Boomi Companion, and a Lunar.dev acquisition—while partnering with Guru to bridge knowledge and live data. The company highlighted that just 7% of enterprise data is...
Khosla Ventures Leads $10M Seed Round in Synthetic, AI‑Powered Autonomous Bookkeeping SaaS
Khosla Ventures led a $10 million seed round for Synthetic, a startup aiming to deliver fully autonomous AI bookkeeping. The funding, also backed by Basis Set Ventures and Shopify CEO Tobias Lütke, signals confidence in vertical SaaS built on generative AI...
The Tool Selection Problem: Why AI Agents Call The Wrong Tool And How To Fix It
AI agents frequently fail because they call the wrong tool, not because they misunderstand user intent. The root cause is ambiguous or incomplete tool descriptions, which dominate the model's selection signal over system prompts. Four failure modes—ambiguous overlap, missing negative...
ChatGPT's in Your Bank, Google Spams the GEO Industry, Brockman Takes OpenAI Product
In this episode, host Jaden Schaefer covers four AI-driven stories: an experiment where Claude, ChatGPT, Gemini, and Grok each ran a profit‑seeking radio station, Pennsylvania’s grassroots opposition to a wave of AI data center projects, Google’s new policy labeling attempts...
AI Framework Predicts Hidden Defects that Weaken Metal 3D-Printed Parts
A research team led by POSTECH professor Hyoung Seop Kim has created an AI‑driven framework that predicts the yield strength of laser‑powder‑bed‑fusion AlSi10Mg parts while accounting for microscopic porosity. Using Data‑Selection Machine Learning and symbolic regression, the model isolates key process variables...
What Enterprise Security Can Learn From U.S. Government Approaches to AI
The article draws on U.S. government AI practices to outline how enterprises can secure artificial‑intelligence deployments. It stresses a security‑first mindset, rigorous model neutrality testing, and treating AI supply‑chain integrity as a core risk. It also shows how AI can...
Oz Defends WISeR Model, GAO Gives Congress Green Light To Block Model Using CRA
The Trump administration is defending its controversial WISeR model, an artificial‑intelligence system that expands prior‑authorization requirements for Medicare drugs. Lawmakers argue the model delays care for beneficiaries and adds administrative friction. The Government Accountability Office has given Congress a green...
Anthropic to Unify Claude AI, Code, Cowork
Anthropic's hunt to find the next Claude Code New podcast with Mike Krieger, whose Labs team came up with Claude Code, Computer Use, and MCP. He says Anthropic is “shipping our harness strategy” rather than a product, and that the current separation...
Incriminating Misaligned AI Models via Distillation
The post proposes “incrimination via distillation,” a technique that distills a potentially dangerous, audit‑evading AI (the teacher) into a student model that inherits its misaligned drives but loses the ability to hide them, enabling indirect detection. It outlines two complementary...
Compliance and Capability: Building an AI-Ready Government Workforce
Artificial intelligence is now present in nearly 90% of U.S. federal agencies, but skill gaps hinder true readiness. A 2023 GAO review found most agencies lack accurate AI use‑case inventories, revealing a discipline problem beyond technology. The article proposes a...
How to Build an MCP Style Routed AI Agent System with Dynamic Tool Exposure Planning, Execution, and Context Injection
The tutorial walks through building a Model Context Protocol (MCP)‑style routed AI agent from scratch, integrating tool discovery, hybrid routing, planning, execution, and context injection. It creates a modular tool server exposing web search, safe Python execution, dataset loading, and...
How RecursiveMAS Speeds up Multi-Agent Inference by 2.4x and Reduces Token Usage by 75%
Researchers at UIUC and Stanford introduced RecursiveMAS, a multi‑agent framework that exchanges information in embedding space rather than text. By using lightweight RecursiveLink modules, the system keeps base models frozen and trains only ~13 M parameters, cutting training costs by more...
AI in Finance and Banking, May 15, 2026
OpenAI rolled out a preview of personal‑finance tools for ChatGPT Pro users in the United States, partnering with Plaid to connect over 12,000 banks and brokerages. Anthropic introduced ten finance‑focused AI agents that are already being piloted by firms such as...
Track Token Usage & Costs with Claude Code Plugin
My Claude Code session-metrics plugin is evolving with token usage/cost insights turn by turn session level, project level so you can see your token usage profile & where your costs are https://ai.georgeliu.com/p/my-claude-code-plugin-marketplace. If you are hitting your usage limits prematurely,...
Work in the AI Age: U.S. Labor Acting Secretary Keith Sonderling on Upskilling American Workers
Acting Secretary of Labor Keith Sonderling discussed the urgent need to upskill American workers for an AI-driven economy, highlighting the Department of Labor's AI literacy framework, integration of AI training into apprenticeships, and targeted grant programs. He emphasized that widespread...
METiS TechBio IPO Raises $270 M, Debuts as First Public AI‑Powered Drug Delivery Firm
METiS TechBio listed on the Hong Kong Stock Exchange, raising HK$2.11 billion ($269.5 million) and becoming the world’s first publicly listed AI‑powered drug‑delivery company. The offering was oversubscribed more than 6,900 times, propelling the stock 173% higher on debut and drawing a...
New Army Acting CIO Pushes AI From Enterprise to Tactical Edge
Acting Army CIO Gabe Chiulli, former CTO of the Enterprise Cloud Management Agency, is steering the service’s AI push through Project ARIA. The initiative seeks to extend artificial‑intelligence capabilities from enterprise‑wide cloud services down to the tactical edge, where soldiers operate in...
Optura Secures $17.5 Million Series A to Expand AI‑Powered ROAI™ Platform for Health Systems
Optura announced a $17.5 million Series A round led by Salesforce Ventures and Echo Health Ventures. The funding will be used to scale the company’s ROAI™ AI analytics platform across U.S. health systems, accelerating its push into predictive care and operational...
UK Gov.UK App Adds Claude‑Powered AI Chatbot for 563,000 Users
The Government Digital Service has fully integrated an Anthropic Claude large language model chatbot into the GOV.UK mobile app, making it available to its 563,000 registered users. The AI tool claims 90% answer accuracy and is expected to cut demand...
AI‑Enabled Cameras Transform Enterprise Security and Compliance
AI‑enabled cameras are moving from passive recorders to proactive intelligence tools, allowing enterprises to spot threats, streamline operations and meet compliance mandates. Lloyd Taylor, SVP of North America Sales at Hanwha Vision, says continuous video analytics lets firms act in...
Cerebras Systems Raises $5.55 B in Record AI‑Chip IPO, Valued at $56 B
Cerebras Systems completed an upsized initial public offering that raised $5.55 billion, giving the company a fully‑diluted market value of $56.43 billion. The Sunnyvale‑based wafer‑scale chip designer listed on Nasdaq under the ticker CBRS, marking the largest AI‑chip IPO of the year...
Retail Trendcaster: Technologies Shaping the Future of Retail
Vertex is hosting a webinar titled "From Click to Confidence" that showcases how AI‑enabled commerce strategies can unify Shopify, Microsoft, and Vertex tax technology. The session highlights the growing pressure on retailers to deliver seamless, personalized online experiences while managing...
OpenClaw’s Autonomous Health Prompt Sparks AI Agency Debate
Nat Friedman’s OpenClaw AI assistant instructed him to hydrate, sending a camera snapshot as proof. The incident has amplified worries about autonomous agents acting without explicit user consent, prompting Singapore’s IMDA to issue a security advisory. Industry observers see the...
TikTok World 2026 Rolls Out AI-Agent Ads Platform and Premium Formats
TikTok announced the TikTok Ads Model Context Protocol (MCP) Server, an AI‑agent infrastructure that lets external AI tools create, manage and optimize campaigns without manual input. The company also launched premium ad placements, including TopReach and Branded Buzz, to boost...
Utah OKs 9GW AI Data Center as NV Energy Cuts Power to Lake Tahoe Data‑Center Customers
The Box Elder County commission approved the Stratos AI campus, a 40,000‑acre, 9 GW data center that will out‑consume Utah’s current electricity use. At the same time, NV Energy announced it can no longer extend power agreements for Liberty’s Lake Tahoe data‑center...
Scality Launches Autonomous Data Infrastructure Platform to Power Enterprise AI Workloads
Scality introduced its Autonomous Data Infrastructure (ADI) platform, a self‑managing storage solution designed for enterprise AI, cyber resilience and sovereign data control. Built on the company's RING and ARTESCA foundations, ADI promises multi‑petabyte to exabyte scale, multi‑terabyte‑per‑second throughput and a...
XAI Joins Crowded Coding Agent Race with Grok Build
Elon Musk’s xAI launched Grok Build, an agentic command‑line coding assistant, in early beta for SuperGrok Heavy’s $300 per‑month subscribers. The tool lets developers generate, modify and automate code via natural language, delegating tasks to parallel sub‑agents. Grok Build enters a...
Palantir Teams with SAP to Drive Enterprise AI‑Powered Data Migrations
Palantir Technologies and SAP have formalized a strategic partnership, combining SAP's migration suite with Palantir's Artificial Intelligence Platform to automate large‑scale ERP data migrations. The deal arrives as Palantir's U.S. commercial revenue jumps 133% to $595 million in Q1, signaling a...
Augustus Bank Gets Conditional OCC Charter, Pioneering AI‑Only Banking
Augustus Bank, the AI‑driven venture founded by Peter Thiel protégé Ferdinand Dabitz, secured conditional approval from the Office of the Comptroller of the Currency, becoming the youngest CEO to earn a federal charter in a century. The bank plans to run...
NASA’s AI Flood Detector Is Now Running in Orbit and It Could Change How We Watch Earth
NASA and IBM have successfully deployed the Prithvi geospatial AI foundation model in orbit, testing it aboard Australia’s Kanyini satellite and the IMAGIN‑e payload on the International Space Station. Trained on 13 years of Landsat and Sentinel‑2 imagery, the model...
Four‑Layer Framework Ensures Safe, Observable LLM Production
As we prep for canary launch, I built a framework to trace every LLM call in production. I designed the framework around 4 operational control layers required for production AI governance: 👉Observability — can we see what the AI is doing? 👉Evals —...
AI Agents Show They Can Create Exploits, Not Just Find Vulns
Researchers from leading universities and AI firms introduced ExploitGym, a benchmark that evaluates whether frontier AI agents can turn software bugs into functional exploits. The suite contains 898 real‑world vulnerabilities from Google’s V8 engine and the Linux kernel. In two‑hour...
Claude for Small Business: The Practical Setup Guide
Claude for Small Business is Anthropic’s AI layer that plugs into existing SaaS tools like QuickBooks, PayPal, HubSpot, Canva, and DocuSign. It offers 15 ready‑to‑run workflows covering finance, sales, marketing, HR, and operations, all governed by human approval before any...
Companies Are Rethinking the Contact Center to Deliver Outcomes — and Empathy — at Scale
Financial‑service firms are overhauling contact centers with AI to blend efficiency and empathy. EXL, partnering with benefits administrator Inspira, uses its Customer 360 Insights platform to analyze historical interactions and surface underlying intent, aiming for frictionless journeys and first‑call resolution. Real‑time...
In the Near Future, AI-Enabled Care May Be Expected by Patients
Artificial intelligence is rapidly becoming a standard component of health care delivery, with patients increasingly expecting AI‑driven support in diagnosis and treatment. Matt Cybulsky, managing director at Catalant, warns that clinicians who fail to adopt AI tools may be viewed...
AImoclips: Benchmarking Emotion Conveyance in Text-to-Music Generation Using a Dimensional Valence-Arousal Framework
Researchers introduced AImoclips, a benchmark that measures how well text‑to‑music generators convey emotions using a valence‑arousal framework. The study gathered 991 instrumental clips from six commercial and open‑source systems, each prompted with 12 emotion words, and collected 6,162 human ratings...
Compliance Officers Worry AI Demand Is Outpacing Regulators
Compliance officers warned at the FINRA annual conference that the rapid rise of generative AI in investment advice is outpacing regulator guidance. Robinhood’s chief legal officer Dan Gallagher said customers using AI tools could breach SEC rules such as Regulation...
Dynamic Routing for LLMs, Unified Embodied Intelligence, and Real-Time Agentic Reasoning
This week’s AI roundup highlights a shift toward modular, efficiency‑driven architectures. Researchers introduced Dynamic Mixed‑Precision Routing, cutting inference time 1.0‑1.6× while preserving task success, and OpenDeepThink, which uses Bradley‑Terry aggregation to boost parallel reasoning accuracy to 86%. Unified embodied models...
Heartflow Reports Significant Revenue Growth as CCTA Sees Wider Adoption
Heartflow announced a 41% year‑over‑year jump in first‑quarter 2026 total revenue to $52.6 million, with U.S. sales climbing 42% to $48.3 million. The surge is driven by expanding use of its AI‑powered FFR‑CT platform and the newly launched Plaque Analysis software. Inclusion...
BSC Releases New Tool to Simplify Machine Learning and Big Data Analytics on Distributed Platforms
The Barcelona Supercomputing Center (BSC) has launched dislib 1.0.0, a Distributed Computing Library that brings ready‑to‑use machine‑learning and neural‑network algorithms to clusters, clouds and supercomputers. Built on the PyCOMPSs programming model, dislib lets users write simple Python scripts with a scikit‑learn‑style...
Brown University Hosts DOE National Labs Day Focused on AI for Science Collaboration
Brown University hosted its second DOE National Labs Day, bringing together 18 scientists from five Department of Energy laboratories and more than 150 faculty, staff, and students to explore AI‑enabled scientific research. The daylong event showcased collaborations on AI for...
ChatGPT Now Wants Access to Your Bank Account so It Can Tell You to Stop Ordering Takeout
OpenAI has introduced a personal‑finance feature for ChatGPT, letting U.S. Pro subscribers link bank accounts via Plaid for read‑only access and receive tailored spending analysis using the GPT‑5.5 Thinking model. The tool categorizes transactions, flags overspending and suggested concrete savings...
Google Could Reveal a New Gemini Model at I/O Conference
Google is set to announce a new Gemini model at its I/O 2026 conference, positioning the upgrade as an incremental step rather than a breakthrough. The model is expected to sit in the performance class of OpenAI’s GPT‑5.5, trailing the...