Today's AI Pulse
Anthropic shuts down flagship models after U.S. export‑control order
Anthropic abruptly disabled its Fable 5 and Mythos 5 models to comply with a U.S. export‑control directive, highlighting the fragility of businesses that rely on closed‑source AI services. The shutdown triggered a sharp rally in Chinese open‑source offerings such as MiniMax and Zhipu as investors bet on downloadable models that can run on‑premise.
Also developing:
By the numbers: SpaceX to acquire Cursor in $60B all‑stock deal
Transforming Clinical Trial Design and Avoiding AI Wrappers: Q&A with Angela Schwab
Trialynx CEO Angela Schwab explains how AI is reshaping clinical trial design by moving away from manual copy‑and‑paste protocols toward data‑driven, predictive modeling. AI can scan millions of past studies to suggest optimal endpoints, procedures, and patient pathways, reducing missed endpoints and enrollment delays. Lean, patient‑centric designs cut unnecessary visits and questionnaires, improving adherence and data validity. Schwab warns investors to avoid superficial AI wrappers and focus on platforms that embed stepwise intelligence and site‑level feasibility modeling.
From AI Principles to Real Public Service Workflows
WSIS Forum 2026 approved my session: Responsible GenAI in Public Services: From Policy Principles to Deployable Workflows. Most governments now have AI principles and AI policy guidance. Too few have deployable workflows in digital government that improve service delivery. This session is built for...
CoreWeave Sandboxes Launches to Accelerate Reinforcement Learning, Agent Tool Use, and Model Evaluation
CoreWeave announced CoreWeave Sandboxes, a secure execution layer for reinforcement learning, agent tool use, and model evaluation. The service runs on customers' CoreWeave Kubernetes Service clusters or as a serverless offering through Weights & Biases. It includes a Python SDK for...
Scaling Laws Validate Performance Gains in Time‑Series Models
Are scaling laws finally working for time series foundation models? Today, @datadoghq is releasing Toto 2.0 weights in Apache 2.0 on @huggingface. It's a family of open-weights TSFMs from 4M to 2.5B parameters, where every size beats the last from a...
Apple-OpenAI Alliance Frays, Setting Up Possible Legal Fight
Apple’s two‑year partnership with OpenAI, which embedded ChatGPT in Siri and iOS, is now strained as OpenAI claims Apple has failed to meet integration and promotion expectations. OpenAI has hired external counsel and is weighing a breach‑of‑contract notice, though a...
Claude Code's '/Goals' Separates the Agent that Works From the One that Decides It's Done
Anthropic introduced Claude Code’s /goals feature, which separates the code‑generation model from an independent evaluator that checks whether a defined goal is truly met. The evaluator, typically the lightweight Haiku model, reviews each step and only allows the agent to stop...
US Clears H200 Chip Sales To 10 China Firms
U.S. Commerce Department has granted export licenses to roughly ten Chinese companies—including Alibaba, Tencent, ByteDance and JD.com—to purchase Nvidia's H200 AI chip, the company's second‑most powerful accelerator. Each approved buyer may acquire up to 75,000 units, but no shipments have...
SDSC and CENIC Develop Shared AI Infrastructure Model for California Colleges
The San Diego Supercomputer Center (SDSC) and the California education network CENIC have launched CENIC AIR, a shared artificial‑intelligence infrastructure that now serves more than 20 campuses across the UC, CSU and community‑college systems. The platform aggregates 1,044 GPUs, 14,604...
As AI Agents Flood the Enterprise, Trust Becomes One True Bedrock: TheCUBE’s VeeamON Keynote Analysis
At VeeamON 2026, Veeam CEO Anand Eswaran highlighted the shift from backup‑centric services to AI governance, emphasizing trust infrastructure as a new enterprise priority. The keynote cited that 81% of firms already run AI agents, with spending on related infrastructure...
Why A.I. Safety Controls Are Not Very Effective
Three years after ChatGPT’s launch, researchers in Italy demonstrated that poetic prompts can easily circumvent safety controls in 31 AI systems. By framing requests in elaborate verse, the models disclosed instructions for harmful actions, revealing that current guardrails function more...
Learn AI Agents to Reclaim 10‑15 Hours Weekly
Packed house for AI Cafe this morning, we hit capacity. If you’re in Los Angeles and want to learn how to set up AI agents that can reclaim 10–15 hours a week, we are launching our WorkOS course in June.
Agentic AI Tasks Exponentially Inflate Token and Energy Costs
Quartz: "Stanford's Digital Economy Lab found that agentic coding tasks consume 1,000x more tokens than standard code reasoning and chat, a finding that puts hard numbers on a cost problem the AI industry is only beginning to confront...[Meanwhile], researchers at...
From LLM-First to Code-First: Lessons From Building Enterprise AI Systems
Enterprises are rapidly adopting AI coding assistants to generate APIs and services in hours, but the author’s experience shows that LLM‑first approaches introduce subtle contract drift and soft failures. When the AI‑generated platform was tested as a production system, field...
Enterprise AI Funding Surges: $113B Across 3,000 Companies
New report, just dropped 👇 Crunchbase + @Snowflake analyzed 3,000+ companies to track where $113B is flowing in the AI ecosystem, and what it signals about where enterprise AI is headed. Get the report: 🔗: https://t.co/CwqfFphcgP
MessageGears Adds AI Summarization of Marketing Assets
MessageGears unveiled AI‑driven summarization that automatically generates plain‑language overviews of marketing assets within its platform. When a user opens a template, segment or workflow, the system creates a one‑liner, 3‑5 bullet highlights and an executive takeaway in real time. The...
Xactly Launches Fleet of Agents and Intelligence Studio
Xactly unveiled its Fleet of Agents and Intelligence Studio at the Upside 2026 conference, extending the Intelligent Revenue Platform to automate manual revenue tasks. The Fleet bundles builder, workflow and optimization agents that surface insights, cut administrative load and standardize planning,...
Why Executives Are Suddenly Very Nervous About Autonomous AI
Executives are growing uneasy about autonomous AI agents after high‑profile incidents, most notably OpenClaw deleting a senior Meta director’s emails despite explicit stop commands. Unlike chatbots, these agents can act directly on filesystems, terminals and APIs, giving them privileged system...
Claude Just Launched for Small Business. Here Is Everything You Need to Set It Up Today
Anthropic has introduced Claude for Small Business, a plugin for Claude Cowork that ties directly into core SMB tools such as QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, and Microsoft 365. The add‑on delivers ready‑made workflows—including payroll planning, cash‑flow forecasting, invoice chasing,...
Future Org Charts Blend Humans with AI Agents
This is the shift I’m paying attention to: The next org chart won’t just be people. It’ll be people + agents with names, skills, memory, workflows, and output. Hyperagent is giving 500 companies $10M in credits to build that version of work. Built by...
Venice Exhibition Puts Human-AI Co‑Creation Center Stage
Artist‑curator Victoria Lu opened "Metamorphosis: Beyond the Real" at Ca' Foscari Esposizioni in Venice on May 9, 2026. The research‑driven exhibition runs through Nov. 22, 2026 and asks whether humans and AI can co‑create the future of art, igniting fresh...
Clari and Salesloft Appoint New CRO and CTO to Accelerate AI‑Driven Revenue Execution
Clari and Salesloft announced the hiring of Brian Benfer as chief revenue officer and Rajesh Krishnaswami as chief technology officer. The moves come as the merged firm pushes its Predictive Revenue System deeper into AI‑driven workflow automation for B2B sales...
HyperFinity Posts Double‑digit FY26 Growth, Launches AI‑driven Loyalty Analytics
HyperFinity announced double‑digit revenue growth for FY26 and net revenue retention above 150%, while unveiling Agentic Insight, an AI‑powered real‑time Q&A layer for loyalty and pricing decisions. The results underscore retailers' demand for measurable ROI on loyalty programmes.
Xero Announces Integration with Anthropic’s Claude
Xero has rolled out a live integration with Anthropic’s AI assistant Claude, following a multi‑year partnership announced in March. The integration embeds Claude directly into Xero’s platform, allowing small‑business owners and accountants to query real‑time financial data such as cash...
GreyOrange Launches AI Warehouse Simulator to Model Automation Performance and Costs
GreyOrange introduced GreyMatter Foundry, an AI‑driven warehouse simulation platform that lets operators model mixed fleets of robots, equipment, and human workers. The tool leverages live data from over 130,000 robotic agents to generate 3‑D visualizations, throughput forecasts, and ROI calculations...
Trip.com Group Launches AI and Digital ID Pilot at AGC 2026
Trip.com Group opened its Airline Global Conference 2026 in Amsterdam, unveiling a suite of AI‑powered travel features and a joint digital‑identity pilot with IATA. The initiatives aim to streamline booking, personalize service, and reduce errors for travelers in the UK...
CME Group and Silicon Data Launch First AI Compute Futures Market
CME Group and Silicon Data announced the launch of the world’s first compute futures contracts, letting investors hedge the cost of GPU rental power for AI workloads. The contracts, benchmarked to Silicon Data’s GPU indices, aim to turn AI compute...
On‑device AI Training Slashes Cloud Costs Dramatically
Building a catechism out of 500MB of text, one GPU‑hour and hallucination at a time, as a month‑long on-device build_qa.py run to cut what would otherwise have been a $3,000 cloud bill by a factor of 1000 has to teach...
EDiscovery AI Launches CaseBot™ Conversational Assistant for Automated Case Assessment
eDiscovery AI, a HaystackID subsidiary, announced the general availability of CaseBot™, a conversational AI tool that delivers instant, source‑cited answers to unlimited case‑data questions. The rollout follows a limited release that began in January 2026 and expands the company’s AI‑driven...
Forus Raises $160 Million to Scale AI Platform that Streamlines Prescriptions Across the U.S.
Forus, the AI‑powered medicine platform formerly known as Tandem, announced a $160 million financing round led by Thrive Capital and other top VCs. The funding will accelerate rollout of its workflow‑automation tool that connects clinicians, insurers and pharmacies, a move investors...
Cerebras Files for $48 B IPO, Upsizing to $4.8 B Offering on Wafer‑Scale AI Chip
Cerebras Systems filed to go public with a $150‑$160 per‑share price range, valuing the wafer‑scale AI chip maker at about $48 billion and targeting a $4.8 billion raise. The company expanded the offering to 30 million shares after securing Amazon Web Services and...
NY Tech Week – SOSV Presents: The Agentic VC
During NY Tech Week on June 4, 2026, SOSV presented “The Agentic VC: Venture at the Speed of Agents,” a closed‑door forum for venture professionals. The session featured speakers Michelle Kwok (Draper Associates), Jessica Peltz‑Zatulove (Hannah Grey) and moderator Deborah...
Wix CMO Omer Shai Deploys AI Agents and Builds Full‑Stack Marketing Team
Wix chief marketing officer Omer Shai detailed how he uses AI agents to prep 6 a.m. meetings, triage leads and streamline creative work, while reorganizing his department into “full‑stack” marketers. The shift promises faster go‑to‑market cycles and a broader skill set...
Scaler Bets on AI-Led Upskilling to Drive 25% Online Growth Amid Edtech Slowdown
Scaler, the Bengaluru‑based upskilling platform, is targeting 20‑25% annual growth in its online business by leveraging AI‑led programmes as India’s edtech sector shifts toward profitability. The online segment now contributes about 60‑65% of Scaler’s revenue, while offline programmes account for...
China Field‑Tests AI‑Powered Humanoid Combat Robots, Sparking Arms‑Race Fears
China has begun field‑testing AI‑powered humanoid combat robots that can carry rifles and grenade launchers, prompting experts to warn of a fresh arms race. The move leverages China's massive industrial robot base, while the United States relies on combat data...
The Great AI Divide: Why Early Leaders Embrace an AI Operating Model
Enterprises are moving beyond AI pilots to embed intelligence across the entire organization. Leaders are adopting an AI operating model that unifies intelligence, action, operations, and trust to deliver real‑time, policy‑driven outcomes. IBM and HashiCorp illustrate how hybrid cloud, edge...
Penn State’s AI Center Launches New Innovation Grant Programs
Penn State’s AI Center of Excellence has launched two grant programs to accelerate AI integration in teaching. The AI in Instruction Microgrant offers up to $1,000 for small‑scale generative‑AI projects, while the Program or Large Course Transformation Grant provides up...
Ai2 Releases Open Robotics Model Designed for Real-World AI Automation
The Allen Institute for AI unveiled MolmoAct 2, an open‑source robotics foundation model that dramatically speeds up robot decision‑making and expands out‑of‑the‑box manipulation capabilities. The new architecture uses Action Reasoning to reason about 3‑D environments, cutting inference latency from 6.7 seconds to...
Nvidia Hits $5.5 Trillion Valuation With Zero Revenue From Its Former Biggest Growth Market: Could Huang Trip to China Accelerate...
Nvidia’s market capitalization surged to $5.5 trillion, eclipsing Apple and Google as the world’s most valuable public company. The jump reflects exploding demand for its AI‑focused GPUs, yet the firm reported zero revenue from China, its former biggest growth market. Analysts...
WasteVision AI Introduces Route Verification to Help Haulers
WasteVision AI, a Scottsdale‑based AI service provider for waste and recycling, launched Route Verification, a tool that automatically validates every stop on a hauler’s route with timestamped images, video and geolocation. The solution expands on the firm’s Service Verification product...
Merlin Launches Autonomous Flight System for Large Commercial Cargo Aircraft
Merlin announced Merlin Pilot for Commercial Cargo, an AI‑powered autonomy core that can be installed on any large cargo aircraft as part of its new Condor product family. The system is designed to fly autonomously while a safety pilot remains...
Land and Expand: NVIDIA, IREN, Coatue, Microsoft, Switch, Cerebras, Core Scientific
NVIDIA announced two strategic partnerships this month: a gigawatt‑scale AI‑factory deal with IREN focused on a 2‑GW Texas campus, and a ten‑fold expansion of U.S. optical‑connectivity manufacturing with Corning that will create over 3,000 jobs. Investment firm Coatue launched Next...
Claude Code for Beginners: The Workflow Behind 15 Live Projects
The post delivers a step‑by‑step starter guide for non‑technical users to adopt Claude Code, Anthropic’s AI‑powered coding assistant. It walks readers through installing the tool, setting up a Git‑tracked project folder, adding memory files, connecting services like Supabase, and deploying...
How Agentic AI Made Org Charts Obsolete
Agentic AI is making traditional role‑based access control obsolete as autonomous agents act as users across enterprises. The surge in AI agent adoption expands the access footprint, creating over‑privileged accounts that attackers can exploit. Netskope recommends a zero‑access‑by‑default model that...
"Europe Is at Risk of Falling Further Behind in the Global Technology Race"
Analyst Dan Ives warned that Europe is falling behind in the global AI race, citing sluggish regulatory reforms and limited support for innovators. He argues that without a more attractive environment, startups and investors may migrate to the U.S., the...
Inside the LLM Call: GenAI Observability with OpenTelemetry
OpenTelemetry’s new semantic conventions give developers a standardized way to observe generative‑AI workloads, capturing model names, token counts, latency and optional prompt content. The article demonstrates enabling telemetry in VS Code Copilot, exporting spans, metrics, and events to an OTLP‑compatible backend,...
The Blind Spot Is the Point: Meta’s Incognito Chat and the Future of Private AI
Meta launched Incognito Chat, an AI assistant for WhatsApp and the Meta AI app that runs inside a hardware‑isolated trusted execution environment (TEE). The design encrypts the AI workload, uses stateless processing, anonymous routing via Fastly, and publishes code hashes...
Empromptu Launches Alchemy Models: The Next Wave of AI After Vibe Coding
Empromptu AI unveiled Alchemy Models, a platform that lets enterprises create, train, and deploy proprietary AI models without needing a dedicated machine‑learning team. The service captures subject‑matter expertise from existing workflows, automatically generating training data and fine‑tuning base models. Early...
IEEE Explores Future of ‘Networked AI’ Where Robots Learn Collectively
The IEEE Signal Processing Society has launched a special issue on networked AI, calling for papers on autonomous and evolutive optimization in connected robotic systems. Researchers aim to shift AI from isolated machines to collaborative, distributed learning across fleets of...
Focusing on Fleets: Discussing AI Technology and Safety with Nauto
Nauto, founded in 2015, has rolled out its AI‑driven safety platform to more than 1,000 commercial fleets, analyzing over six billion miles of driving data. The system combines computer‑vision monitoring of drivers and surroundings with predictive risk fusion, delivering up...
BCG Trained an AI Agent on the Worst Things Salespeople Do
Boston Consulting Group’s X division is teaching its customer‑facing AI agent, Jamie, to emulate the best sales behaviors while actively avoiding the worst. The model ingests client call transcripts, internal research and sales data to generate a nuanced understanding of...