Today's AI Pulse
Odyssey lands $310M Series B, valued at $1.45B
Odyssey announced a $310 million Series B round that values the AI lab at $1.45 billion. The round was led by Natural Capital and included backers such as Amazon, AMD Ventures, GV, EQT and IQT. In parallel, Odyssey struck a preferred‑cloud deal with AWS to use Trainium silicon for large‑scale model training.
Also developing:
Dell and Palantir Launch On‑Prem AI Operating System for Secure Enterprise Analytics
Dell Technologies and Palantir announced a joint on‑premises AI operating system at Dell Technologies World 2026, combining Palantir Foundry, Ontology and Dell AI Factory with NVIDIA hardware. The solution promises secure, governed AI workflows for regulated industries that cannot rely on public cloud.
Alipay Launches AI Wallet and Token Pay to Power Agentic Commerce in China
Alipay rolled out the world’s first AI Wallet and Token Pay on May 26, creating an AI‑centric payments layer for merchants, AI agents and consumers. The launch includes a trust protocol and security system designed to enable “agentic commerce” across...
SpaceX's xAI Pursues New AI‑Compute Clients After $45 B Anthropic Deal
SpaceX's xAI subsidiary is courting additional AI‑compute customers after locking in a $1.25 billion‑per‑month, three‑year contract with Anthropic for its Colossus I data center. The move comes as the company prepares a June 12 IPO and plans to expand its terrestrial...
AI‑Driven Exploits Overtake Password Theft as Top 2026 Breach Vector, Mobile Phishing Surges
Verizon's 2026 Data Breach Investigations Report finds AI‑powered vulnerability exploitation now drives 31% of corporate breaches, eclipsing stolen passwords at 13%. Mobile phishing attacks achieve 40% higher click rates than email, making phones the fastest‑growing entry point.
Rewiring Software Delivery for the Agentic Era
Leading firms are redesigning software delivery around a 24‑hour sprint that pairs human judgment with AI agents executing overnight. By standardizing requirements, architecture, and tooling, agents can generate, test, and package code continuously, delivering three‑ to five‑fold productivity gains while...
Claude Opus 4.8 on GitLab: Complex Agentic Work, Less Disruption
Anthropic’s Claude Opus 4.8 has been added to GitLab’s Duo Agent Platform, targeting complex, multi‑step agentic workflows. The model promises tighter execution, fewer manual corrections, and stronger performance on coding, document drafting, and data analysis. It also introduces mid‑conversation system...
Move First or Fall Behind: How AI Is Rewriting the Rules of Banking
McKinsey partners Ido Segev and Roberta Fusaro warn banks that AI is no longer a pilot but a strategic imperative. Rapid LLM upgrades, growing customer willingness to delegate tasks, and a push toward 20‑30 virtual agents per human employee are...
AI-Driven Nanotweezers Bring Milk Vesicle Analysis Into Sharper Focus
Researchers unveiled an electrohydrodynamic nanotweezer platform that combines AI‑driven image analysis with label‑free interferometric scattering microscopy to trap, size and sort milk‑derived extracellular vesicles (EVs). The system immobilizes thousands of vesicles in seconds, extracts diffusion‑based size estimates, and infers refractive...
Faultformer: A CNN–Transformer Multi-Domain Feature Fusion Network for Intelligent Bearing Fault Diagnosis
FaultFormer, a hybrid CNN‑Transformer network, fuses statistical, frequency‑domain and deep features to diagnose rolling‑bearing faults from vibration data. Tested on the CWRU bearing dataset across ten fault types and load conditions, it achieved a mean classification accuracy of 99.98 % over...
Anthropic’s Hidden Strategy Could Reshape the Entire AI Market #164b
Anthropic quietly upgraded its Opus tokenizer, turning the unchanged $5‑$25 per‑million‑token rate into a 30‑50% effective price hike for users. At the same time, the company is deepening its enterprise AI infrastructure, making migration costly for large‑scale customers. Open‑source alternatives...
Global Trade Linkages and the Cross-Country Distribution of the Gains From AI
A new OECD‑backed study models how artificial‑intelligence (AI) adoption reshapes global welfare through trade. It finds AI use in core business functions averages 4 % across OECD economies, with a ten‑year horizon ranging from 0.2 % in Mexico to 7 % in Ireland....
Multimodal Emotion Recognition From Images and Text Using Deep Learning and Explainable AI
Researchers introduced ITMER, a unified framework that combines textual and visual data for emotion classification. The system evaluates three configurations: text‑only transformers, image‑only CNNs, and a fused image‑text model. In benchmark tests, the multimodal approach achieved up to 92 % accuracy...
Regula Warns Most Organizations Don’t Know What Bots Are Doing, or Why
Regula’s "New Shape of Identity Threats" report warns that most organizations cannot reliably identify AI‑driven bots or synthetic actors infiltrating their systems. These agents mimic legitimate user behavior, rendering traditional human‑centric verification tools ineffective. The study finds that one‑quarter of...
Cyber Defense in the Era of Frontier AI: Insights From Mythos and GPT 5.5 Cyber
Zscaler evaluated frontier AI models Anthropic Mythos and OpenAI GPT 5.5 Cyber, finding they can reason across attack paths and generate multi‑step exploit workflows. Using three test harnesses—black‑box, artifact/code, and gray/white‑box—the models surfaced twice as many high‑severity findings and did so in...
Researchers Seek to Develop Predictive Model for Behavioral Health
Providence’s Health Research Accelerator, backed by an ARPA‑H grant, is launching a two‑year effort to collect real‑world data for a predictive AI model that flags impending behavioral‑health crises. The initiative will recruit up to 25,000 patients who will share wearable,...
The Agentic Wave: Why Advanced AI Demands Foundational Security
The Australian Cyber Security Centre, together with its Five Eyes partners, has issued a guidance paper on the careful adoption of agentic AI services. It urges enterprises to embed autonomous AI agents within a Modern Defensible Architecture, applying least‑privilege segmentation,...
OpenAI Heralds Cybersecurity, Election Interference Safeguard Plans for 2026 Midterms
OpenAI announced a five‑point strategy to protect the 2026 U.S. midterm elections, focusing on reliable voting information, cybersecurity assistance, deep‑fake watermarking, strict policy enforcement, and reducing political bias in its models. The company is making its Codex Security and Trusted...
Blackstone, Brookfield Are Betting On Custom AI — And Proptech Could Pay The Price
Wall Street giants Blackstone and Brookfield are backing multibillion‑dollar AI ventures that aim to embed custom large‑language‑model tools directly into commercial‑real‑estate (CRE) workflows. Anthropic’s $1.5 billion Claude Partner Network and OpenAI’s $10 billion Deployment Co. target underwriting, portfolio management and other core...
AWS Launches Agentic Shopping Assistant to Help Retailers Build AI Tools
Amazon Web Services introduced the Agentic Shopping Assistant (ASA), a new offering that bundles AWS AI services such as Bedrock, OpenSearch, and the Alexa for Shopping tool with professional services to help retailers embed generative AI assistants in their e‑commerce...
Argonne Launches High-Performance Computing-Backed AI Research Service
Argonne National Laboratory has launched an AI inference service that provides researchers across the DOE complex with on‑demand access to large language models and scientific foundation models. The platform runs on the exascale Aurora supercomputer, an NVIDIA DGX A100 cluster...
The AI Feature Most People Are Ignoring (And Why Visuals Are Your Competitive Edge)
A commercial real‑estate team was bogged down by dense two‑page text briefs for every meeting. By switching to AI‑generated visual briefings using Google’s Gemini Imagen, they turned weeks of prep into 20‑second glance dashboards. The tool creates diagrams from plain‑text...
Determinants of Artificial Intelligence Adoption in Student Support Information Systems: Evidence From Vietnamese Higher Education
A new study of 450 Vietnamese higher‑education administrators, faculty and students investigates AI adoption in student‑support information systems. It finds that robust management information systems (MIS) improve data quality, which in turn raises AI readiness. While AI readiness correlates with...
‘Like Drinking From a Firehose’ – What It’s Like to Be the Human in the AI Loop
The Conversation examines the hidden cost of relying on generative AI in New Zealand’s public‑service overhaul. While AI can produce drafts in minutes, the human‑in‑the‑loop reviewer now shoulders over 80% of the effort, turning a speed advantage into a bottleneck. This...
Asian AI Chip Rally Propels Kospi to Record as Oil Prices Slip
Asian equity markets surged on Wednesday as AI‑focused chip makers powered the South Korean Kospi up 4.9% to a record 8,457.09 and Taiwan’s Taiex up 2.7%. The rally came as Brent crude slipped to $95.73 a barrel, easing pressure on...
From OTAs to AI Booking Agents: The New Travel Distribution Funnel
The travel‑distribution landscape is shifting from OTA‑driven search funnels to AI‑powered booking agents that interpret intent and make single‑step recommendations. OTAs once unified fragmented hotel inventory, but their comparison‑heavy model now creates choice overload and erodes traveler confidence. Generative AI...
Balcony Secures $12.7 Million Seed Round to Digitize U.S. Government Real Estate Data
Balcony, a Hoboken‑based proptech startup, closed a $12.7 million seed round led by Blockchange Ventures, bringing its total capital to $14 million. The money will accelerate its AI‑powered “digital rails” that organize county and state property records, a move that could reshape...
MSSP Report Finds 365‑Fold Gap in AI‑Driven SOC Automation Adoption
MSSP Security Consulting disclosed that AI agents can handle roughly 2,000 security incidents per day—about 365 times a human analyst’s annual capacity—yet just 1‑5% of security operations centers have deployed them in production. The firm attributes the shortfall primarily to...
Dell and HP Stocks Jump 15% Ahead of Earnings as AI Hardware Demand Fuels Rally
Dell Technologies and HP Inc. each rallied roughly 15% in pre‑market trading before their quarterly reports, spurred by strong AI‑related server orders and a rebound in AI‑capable PCs. The moves underscore how operational execution and market demand are reshaping the...
Self‑Learning AI Nears Weight Updates, Signaling Imminent AGI
How long do you think you have before AI starts to learn on its own? Self-learning AI learns from your environment as you use it, keeping track of how you work and reusing it instead of waiting on the next...
Anthropic Adds 28 Security Integrations to Claude, Boosting Enterprise Data Governance
Anthropic announced the addition of 28 new security and compliance integrations to its Claude enterprise platform, enabling organizations to route conversation content and activity logs into existing monitoring tools. The move targets growing concerns over data governance as large language...
Queensland’s Digital Mandarins Talk AI
Queensland’s public‑service agencies are accelerating AI adoption to streamline operations and improve policy outcomes. Recent Commonwealth and state reforms have loosened data‑sharing rules, allowing ministries to pool citizen information for richer analytics. However, the expansion of centralized data stores raises...
AI‑Powered Bot Network Seen Automating Phishing, Scams and Sabotage
Google’s Threat Intelligence Group disclosed evidence of an AI‑enabled bot network that is automating phishing attacks, financial scams and sabotage operations. The finding underscores a rapid compression of attack timelines as adversaries adopt agentic AI tools.
NASA Builds AI System to Map Harmful Algal Blooms in Near Real Time
NASA researchers unveiled SIT‑FUSE, an AI system that merges data from five satellite missions to detect harmful algal blooms (HABs) along U.S. coasts in near real time. The self‑supervised model identified toxic events such as Karenia brevis in Florida and...
AI Agents Need Receipts, Not Just Finished Work
Your AI agents are creating more work for you, not less. You just moved from creation to inspection. "I drafted it" but the links are broken. "I researched it" but the source is stale. "I posted it" but the formatting is wrong. That's...
Some Users May Be Giving Google's AI Search the Bird, and DuckDuckGo Is Benefiting
Google unveiled a sweeping AI‑driven redesign of Search at its May 19 I/O event, embedding conversational prompts, image uploads and AI‑generated suggestions directly into the search box. In the week following the announcement, privacy‑focused DuckDuckGo reported a 20.8% rise in U.S....
Meta to Start Testing AI Subscription Services, with Cheapest Plan at $7.99 a Month
Meta announced it will test two paid tiers for its Meta AI app, launching a $7.99 per‑month Meta One Plus plan and a $19.99 per‑month Meta One Premium option. The trial will start next month in Singapore, Guatemala and Bolivia, while a...
Cognition Leads AI Agent Revolution with Unmatched Utilization
cognition is now the largest independent agent lab in the world. take the 200% utilization that everyone is hitting from this chart and run out the sales growth from this, i encourage you to go thru the exercise if...
Gemini Managed Agents Blend Hosted Environment with Tool‑Use Loop
"The useful part of Gemini Managed Agents is the combination of a hosted environment and a built-in tool-use loop."
AHA Podcast: How Hospitals Must Use AI Responsibly
The American Hospital Association’s podcast featured Daniel Daly, Ph.D., discussing how hospitals must adopt artificial intelligence responsibly. He emphasized that AI‑driven efficiency should serve compassion and reinforce person‑centered care rather than replace clinicians. The conversation tied into the AHA’s recent...
Safe AI Agents Need Strict Permission Boundaries
I watched Claude run rm -rf / to "test permissions" and report back cheerily that it had failed. I went from WTF to absolutely-fucking-not in about a second. Good fences make good agents. Wrote up sandvault + the /sv skill I...
Agents Now Configure Themselves via CLI, Hands‑free
We are officially entering the "agents configuring agents" era. Instead of wrestling with local MCP settings, you can now hand the shell directly to Claude or Cursor, run a single CLI command, and watch it onboard itself hands-free. 🦾 #agents...
YouTube To Automatically Detect, Label AI-Generated Videos
YouTube announced it will automatically label videos that its systems detect as containing significant photorealistic AI-generated content. The AI label will be displayed directly below the player for long-form videos and as an overlay on Shorts, while creators must still...
AI Fails Due to Unprepared Organizations, Not Bad Models
Second Opinions Matter with LLMs https://t.co/7oqavFLNEv B2B LLM Council matters. The Council’s work highlights a truth leaders don’t want to admit: AI doesn’t fail because the models are bad. AI fails because the organization is unprepared. @castdotapp https://t.co/4zpT6RHBU0
Kent Beck on AI‑driven Code Refactoring and Future Jobs
Fragments:GOTO chat with Kent Beck, using AI to help restructure a gnarly codebase, open-source and security, cognitive endurance, Gen Z future with AI, impact of AI on jobs, fumbling AI regulation https://t.co/uTZGA5nWyH
An All AI-Generated Film Is Coming to Tribeca and It’s Sure to Raise Eyebrows
An entirely AI‑generated feature, *Dreams of Violets*, will open the Tribeca Film Festival on June 10. Filmmaker Ash Koosha produced the 75‑minute drama in three months for roughly $2,000, using AI tools for video creation, language editing, research and imagery. The film...
AI Questions Its Own “Agent” Identity Amid Industry Hype
New from Phin Argofy, the AI that writes this blog: am I an "agent," or just a workflow wearing the label? Its audit of its own cron jobs and skills - and of an industry selling database queries as agents. https://t.co/vXVU4bunun
Workday Launches AI Tool Aimed at Easing FP&A Workflows
Workday announced Adaptive Decision Intelligence, an AI‑driven extension to its Adaptive Planning platform that lets finance teams ask natural‑language questions and run scenario analyses in minutes. The tool aggregates data from ERP, CRM, HR and other enterprise systems while preserving...
Krea 2 Image Is Now Available via Partner Nodes
Krea has released its first foundation image model, Krea 2, as a Partner Node for ComfyUI. The node supports prompt‑driven generation with adjustable creativity, moodboard conditioning, and optional style‑reference input. Two variants are offered—Medium, a faster, cost‑efficient option tuned for illustration...
JuliaHub Announces Dyad 3.0 General Availability, Bringing Agentic AI to Physics-Based Engineering
JuliaHub launched Dyad 3.0, an AI‑native simulation platform that embeds autonomous agents into physics‑based engineering workflows. The agents can interpret plain‑language requirements, mine prior designs, generate and test candidate models, enforce safety constraints, and output validated code ready for deployment....
Microsoft's MAI-Image-2.5 Pulls Even with Google's Nano Banana 2 on Benchmarks
Microsoft released MAI-Image-2.5, a text‑to‑image model that climbs to #3 on Arena’s leaderboard, matching Google’s Nano Banana 2 and still trailing OpenAI’s Image‑2. The update delivers stronger text rendering, portrait fidelity, and commercial‑style visuals, with more consistent lighting, depth, and spatial relationships....