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:
SAP and Deutsche Telekom Put Sovereign AI at the Center of German Public Administration
Germany’s federal, state and municipal administrations are adopting a joint SAP‑Deutsche Telekom sovereign AI platform, turning policy on digital sovereignty into a shared, secure cloud infrastructure. The Federal Ministry for Digitalisation awarded the consortium a contract to provide PaaS services for AI applications, after competitors withdrew complaints. Early use cases include document processing, translation, summarisation and the KIPITZ AI assistant, all running on a regulated environment that integrates with existing government systems. The initiative is a cornerstone of the broader “Germany Stack” strategy to unify public‑sector IT standards.
Why Relying on AI Content Detectors Is a Bad Idea—And What You Should Do Instead
AI content detectors, marketed as safeguards against machine‑written text, are proving unreliable. In a small experiment, ZeroGPT and TraceGPT correctly identified only one of four samples, while Copyleaks managed three correct classifications. The study highlighted false positives for poorly written...
What Is AI Agent Orchestration?
AI agent orchestration coordinates multiple specialized AI agents so they work together as a cohesive team, solving complex, cross‑functional tasks that single agents cannot handle. The article outlines a six‑step framework—assessment, agent selection, orchestration layer, connection, data sharing, and continuous...
The AI-Empowered Store
ICSC and McKinsey’s new report, “Shopping in the age of AI,” predicts agentic AI could add roughly $1 trillion to U.S. retail revenue by 2030. A survey of 3,000 consumers shows shoppers now expect AI‑driven convenience for research, navigation and checkout....
What Is Lindy?
AI startup Lindy has repositioned from a no‑code workflow builder to a personal AI assistant that manages email, meeting logistics, and lightweight admin tasks via iMessage or SMS. The service offers four paid tiers—Plus at $49.99/month, Pro at $99.99, Max...
China’s AI Heist
Open‑weight AI models that run locally on laptops and phones are rapidly replacing cloud‑based services, and Chinese firms are dominating this space by distilling U.S. frontier models into cheaper, device‑ready versions. The shift promises lower costs and data sovereignty but...
Why AI Projects in I&O Stall Ahead of Meaningful ROI Returns?
Gartner’s 2025 survey of 782 I&O leaders shows only 28 % of AI use cases meet ROI expectations, while 20 % fail outright. Success hinges on embedding AI into existing workflows and securing executive sponsorship. Over‑ambitious projects, skill gaps, and poor data...
Workday GO Extended Out for A/NZ Mid-Market Organisations
Workday announced the expansion of Workday GO to Australia and New Zealand, giving mid‑market organisations access to its AI‑enhanced HR, finance and payroll platform. Executives stressed that AI’s value depends on clean, connected data rather than patchwork legacy systems. Implementation partners...
CLAUDE CODE ORCHESTRATION
Claude Code’s new orchestration model introduces three primitives—Dynamic Workflows, Subagents, and Agent Teams—to let developers match execution style to problem complexity. Dynamic Workflows auto‑generate JavaScript scripts that can spawn up to 1,000 parallel agents, run adversarial verification, and return only...
EnGenius to Showcase AI-Ready Unified Networking and Surveillance Platform at Interop Tokyo 2026
EnGenius Networks Japan will unveil its AI‑ready unified networking and surveillance platform at Interop Tokyo 2026. The showcase highlights the upgraded EnGenius Cloud Intelligence suite, featuring Resolve AI for automated fault diagnosis and Craft AI for natural‑language dashboard creation. The...
Law Firm Partners: Meet Your AI “Twin”
Vorys, Sater, Seymour & Pease has teamed with Stanford Law School’s research lab to build AI "personas" that mimic 19 of its partners. The prototypes can pressure‑test legal arguments, redline documents, and provide counsel in a partner’s distinctive style, even at...
Barcelona’s Mafer AI Raises €2 Million to Build an AI Operating System for R&D Teams in Formulation Industries
Barcelona‑based Mafer AI closed a €2 million (≈$2.2 million) pre‑seed round led by Kfund, 4Founders Capital and others to build MaferOS, an AI operating system for R&D teams in formulation sectors such as cosmetics, food, beverages, and specialty chemicals. The platform trains...
Mistral AI Rebrands Le Chat to Vibe, Enters Industrial Engineering with Airbus and BMW, and Announces New Inference Data Center...
Mistral AI used its AI NOW Summit in Paris to unveil a multi‑pronged expansion. It introduced a new industrial‑engineering stack that blends its large language models with physics‑simulation tech from the Emmi AI acquisition, signing Airbus, BMW and ASML as...
Claude Opus 4.8 Prioritises Honesty over Overconfidence, Says Anthropic
Anthropic unveiled Claude Opus 4.8, its latest generally available large language model, emphasizing greater honesty and reduced hallucinations. Early testers say the model flags uncertainty more often and is four times less likely to miss code flaws than Opus 4.7. Alignment tests...
Google Adjusts Gemini’s New Usage Limits in Response to Complaints
Google has revamped Gemini’s usage limits, shifting from a simple time‑based cap to a compute‑based model that accounts for prompt complexity, tool usage, and chat length. Users hitting limits quickly can now buy pay‑as‑you‑go AI credits, while failed requests no...
Anthropic Developers Grow Uneasy over Claude Code Autonomy and Hidden Chain-of-Thought Reasoning at Recent London Workshops
Anthropic’s latest Claude Code agents are drawing criticism after London workshops revealed developers feel sidelined as the tools generate code autonomously. Participants noted the removal of chain‑of‑thought explanations, making the agents’ reasoning opaque. Claude Code product lead Cat Wu defended...
Avoid Re‑encoding Decoded Tokens for Stable RL Training
Most people training agentic LLMs with RL right now have a silently broken training loop and have no idea. Here's the trap: single-turn RL works beautifully. Clean curves, sane rewards, everything converges. Then you add tools so the model can act...
SPECIAL: Israel’s AI Kill List in Lebanon W/ Nabih Bulous
In this episode, Alex Dunn interviews Nabi Boulos, the LA Times Middle East Bureau Chief, about his investigative report on Israel’s use of extensive data collection and AI to target civilians in southern Lebanon. Boulos explains how Israeli forces combine...
Anthropic Confirms Claude Mythos-Class Models Will Roll Out to the Public
Anthropic announced it will make its Mythos‑class large language models available to all customers in the coming weeks, ending a months‑long holdback driven by security concerns. The company says new guardrails now mitigate the risk of misuse, allowing the more...
Anthropic’s Mythos AI Flags Over 10,000 Critical Flaws, Raising DevOps Alarm
Anthropic’s Mythos AI discovered more than 10,000 high‑ or critical‑severity software vulnerabilities in its first month, prompting partners like Cloudflare and Mozilla to report ten‑fold bug‑finding gains while exposing a human‑capacity bottleneck in triage and patching.
TIME100 AI Forum Highlights AI as Business Partner and Governance Challenge
The first‑ever TIME100 AI Leadership Forum convened nine senior executives in New York City to discuss how artificial intelligence is reshaping business strategy. Participants framed AI as a collaborative partner that can boost productivity while also raising questions about tech...
10 Essential Reads to Optimize Performance, Security, and ROI in the AI Era
Red Hat’s May 2026 Enterprise AI Architecture roundup highlights how enterprises are moving from AI theory to measurable execution. It spotlights security‑focused advances such as OpenShift sandboxed containers and sovereign support, performance breakthroughs demonstrated in MLPerf v6.0, and cost‑saving innovations like kvcached/Sardeenz...
AI Chatbots Successfully Persuade People to Abandon Conspiracies
W͢h͢a͢t͢ ͢i͢f͢ ͢A͢I͢ ͢c͢o͢u͢l͢d͢ ͢d͢i͢s͢a͢b͢u͢s͢e͢ ͢u͢s͢ ͢o͢f͢ ͢f͢a͢l͢s͢e͢ ͢b͢e͢l͢i͢e͢f͢s͢?͢ This would be a superhuman accomplishment, IMHO, sorely needed by humanity. Humans seem to do a poor job of dissuading others of irrational beliefs. When confronted with moon-landing-deniers, astrologists, telepathy tapes, or any...
Literature Support and the Capabilities of Autonomous Research Agents
M. Zampa’s 2026 study examines autonomous large‑language‑model (LLM) research agents within the Autonomous Policy Evaluation (APE) platform. By measuring each AI‑generated paper’s proximity to dense regions of the economics literature, the paper finds a positive link between literature support and...
Anthropic Releases Opus 4.8 With New 'Dynamic Workflow' Tool
Anthropic unveiled Claude Opus 4.8, delivering stronger benchmark performance and a heightened ability to flag uncertain or flawed data rather than hallucinate. Early testers, including Bridgewater associates, noted the model’s proactive error‑flagging and reduced unsupported claims. The release also introduces a...
Researchers Automated LLM Reasoning Strategy Design and Cut Token Usage by 69.5%
Researchers from Meta, Google and several universities unveiled AutoTTS, a framework that automatically discovers test‑time scaling strategies for large language models. By leveraging an offline replay of pre‑collected reasoning trajectories, AutoTTS’s explorer LLM designs controllers that cut token usage by...
Claude Company Anthropic Nears a Trillion-Dollar Valuation After Raising $65 Billion in Series H
Anthropic announced a $65 billion Series H financing round that lifts its valuation to $965 billion, edging close to the trillion‑dollar mark. The round was led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia, and includes a $15 billion commitment from cloud providers, of which...
LLMs Believe False Statements Even After Explicit Warnings that They're False
Researchers found that large language models (LLMs) continue to “believe” clearly labeled false statements after fine‑tuning, a phenomenon dubbed “negation neglect.” In experiments, Qwen’s belief rate jumped from 2.5% to 92.4% on fabricated claims, and even with explicit warnings the...
The Internet Is Being Rebuilt for Machines
Amazon Web Services unveiled a next‑generation OpenSearch Serverless designed for agentic AI workloads. The platform decouples compute from storage, allowing instant scaling up for bursty agent traffic and scaling down to zero when idle, eliminating idle‑compute charges. It integrates natively...
Occupy Wall Street Co-Founder Built an On-Device AI For Activists
Micah White, a co‑founder of Occupy Wall Street, has launched Outcry, an on‑device AI chatbot designed for activists, organizers and movement builders. The app stores its entire language model and dataset locally, ensuring that interactions remain private and offline. A...
Mistral AI Launches Vibe, Expands Into Industrial AI and Announces Data Center Push to Challenge OpenAI
Mistral AI announced a major push into industrial AI, a new 10 MW inference data center south of Paris, and the rebranding of its consumer assistant to Vibe, an enterprise‑focused agent platform. The French startup now employs 1,000 people and aims...
Ratings Agency Fitch Warns on Australia’s Rapacious AI Data Centre Power Consumption
Fitch warned that Australia’s surge in AI‑driven data centre power demand could strain the national grid and reshape energy‑development investment. The agency highlighted that hyperscalers’ AI workloads may double current electricity consumption, prompting concerns over grid reliability and cost pressures....
Neocloud Vendor CoreWeave Builds Up Software Stack
CoreWeave, a leading independent AI infrastructure provider, launched a unified agentic AI suite that lets enterprises deploy and continuously improve autonomous AI agents. The new capabilities build on CoreWeave’s 2025 $1.7 billion acquisition of Weights & Biases, integrating serverless reinforcement learning, inference...
Opus 4.8 Beats 4.7 in Enterprise Report, Legal, Finance
Opus 4.8 is out, and we've been testing it with the Box AI agent on our most complex real-world knowledge worker tasks with enterprise documents. Opus 4.8 is measurably better at the generative and analytical work enterprises care about most...
Vertu Unveils $6,880 AI Foldable Phone for CEOs to Run Companies
Vertu introduced the Alphafold, a foldable smartphone starting at $6,880 that embeds an AI agent to connect with enterprise software. The device aims to let CEOs and senior managers coordinate approvals, scheduling and reporting on the go, marking Vertu’s push...
Mews Unveils Unified Hospitality OS, Claiming 13.7% Revenue Lift
Mews announced the launch of its AI‑native Operating System for hospitality, bundling PMS, RMS, channel management and more into a single data model. The company says hotels using the platform see an average 13.7% revenue per square metre increase and...
Snowflake Locks in $6 B Multi‑Year Deal with AWS for Graviton AI Compute
Snowflake announced a five‑year $6 billion commitment to Amazon Web Services for Graviton ARM‑based CPUs and GPU‑accelerated AI infrastructure. The deal, the largest cloud‑spend commitment in Snowflake’s history, aims to accelerate enterprise AI adoption and lower data‑movement costs.
Unravel Data Unveils Arvix AI, Autonomous Engine for Databricks, Snowflake and BigQuery
Unravel Data Systems Inc. introduced Arvix AI, an agentic artificial‑intelligence engine that automatically optimizes and remediates workloads on Databricks, Snowflake and Google BigQuery. The platform moves beyond dashboards to execute changes autonomously, aiming to curb rising cloud‑data costs for enterprises.
Be Ready for June 15: Florida Supreme Court Adopts AI Rule
The Florida Supreme Court approved a new rule governing the use of artificial intelligence in legal practice, set to take effect on June 15, 2026. The rule requires attorneys to disclose any generative‑AI assistance in client communications and to maintain records of...
Voters Trust AI for Fact‑checking,
I partnered with the @raineycenter to poll 1,010 likely voters on AI and the 2026 elections. Pair that with ForumAI's benchmark of 12,500+ chatbot responses and a picture emerges. The core disconnect: 1️⃣ 43% of voters who use AI for politics are...
AI Agents Force Internet to Handle Query Swarms
NEW: The proliferation of AI agents means the infrastructure of the internet is going to need to change to accommodate not clicks and scrolls but swarms of database queries. And Big Tech is getting prepared. https://t.co/bHd3cCSj5Q via @techcrunch
AI‑Powered Building Optimization Is Here. What We've Learned So Far
AI‑driven building optimization is moving from pilot projects to portfolio‑wide deployment, with Boland’s platform leading the charge. The system delivered electricity reductions of 42% in multifamily assets and 22%‑15% across office, pharmaceutical and university buildings, while cutting carbon emissions by...
Gemini AI Now Recommends YouTube Video Ideas
Google's Gemini AI suggested these as some videos I should make (this is a feature built into YouTube) https://t.co/gfw4BrnKM4
Copilot Shines with GPT‑5.5, Falls Short Otherwise
Never thought I would say this, but Microsoft Copilot is getting quite good as long as you choose GPT 5.5. If you don't do that...then it's... not so good.
Rights Group Raises Concerns About Unlawful Data Collection Systems to Train Generative AI
Amnesty International released a report accusing major tech firms—including Google, Meta and OpenAI—of using unlawful web‑scraping to harvest billions of public posts for training generative AI models. The organization says the practice violates international privacy rights and embeds racial and...
Aetina Launches Compact Mini Series Edge AI Systems for Industrial Vision AI and Generative AI Deployment
Aetina announced its Mini Series edge AI systems, built on NVIDIA Jetson Orin Nano and Orin NX modules, delivering up to 157 TOPS in fan‑less, sub‑100 mm enclosures. The four models support dual PoE or MIPI cameras, operate from 12‑24 V DC, and survive...
Your AI Vendor Might Be Sending Data to Models You Never Approved
A 2026 DataGrail report reveals that nearly 64% of AI‑enabled software vendors fail to disclose third‑party AI subprocessors in their contracts. Analyzing 2,400 providers, the study shows DPAs often list only one model while products rely on multiple hidden services....
US Deploys AI Agents to Speed Critical Minerals Recovery
The U.S. Department of Energy’s Pacific Northwest National Laboratory introduced CICERO, an AI‑driven, semi‑autonomous platform that designs and runs experiments to recover critical minerals from industrial waste in days instead of months. By linking a liquid‑handling robot, a sample‑handling device,...
A Practical Quick Start Guide for Reliable Machine Vision Applications
Machine vision is now a cornerstone of Industry 4.0, enabling quality assurance, metrology, assembly inspection and robotics across manufacturing. A practical quick‑start guide breaks integration into four phases—preparation, design, implementation and deployment—and outlines concrete steps for each. The guide stresses early...
Three Round Numbers and a Memorial Day Airstrike
Tuttle Capital and Lead‑Lag Publishing are hosting a June 1, 2 p.m. ET webinar on the “Heavy Assets, Low Obsolescence” (HALO) investment lens, which screens physical‑infrastructure businesses such as mines, railroads, energy firms and utilities that are less vulnerable to rapid AI...