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:
Google's New Gemma 4 12B Model Is Designed to Run on Any Laptop with 16GB of RAM
Google unveiled Gemma 4 12B, a 12‑billion‑parameter generative AI model that can run on consumer laptops equipped with 16 GB of RAM, eliminating the need for expensive AI accelerators. The model bridges the gap between the lightweight mobile‑optimized Gemma variants and the larger 26B‑parameter models, delivering near‑equivalent performance with roughly half the memory footprint. It introduces Multi‑Token Prediction (MTP) for faster token generation and a streamlined multimodal embedding that removes separate vision and audio encoders. Weights of about 18 GB are available under an Apache 2.0 license via Kaggle and Hugging Face, and the model can be accessed through tools like LM Studio and Google AI Edge Gallery.
Clinical and Economic Evaluation of Autonomous Artificial Intelligence in Musculoskeletal Therapeutics: Flok Health's Series A and the Shifting Competitive Landscape
Flok Health, a Cambridge‑based digital health firm, closed an oversubscribed $12.5 M Series A led by AlbionVC to scale its autonomous AI‑driven physiotherapy clinic across the NHS and into international markets. The platform, a CQC‑registered Class IIa medical device, can triage, diagnose, treat...
Salesforce Acquires Contentful, Rolls Out New AI Agents for Marketers
Salesforce signed a definitive agreement to acquire Contentful, the composable content platform used by over 4,800 brands, and will embed its headless CMS into the Customer 360 suite. The transaction, slated to close in the third quarter of fiscal 2027,...
AI’s Productivity Is Finally Hitting the Real Economy
A Federal Reserve Bank of St. Louis report finds generative‑AI use in U.S. businesses jumped 10% year‑over‑year, with 54.6% of working‑age adults using AI by August 2025. Survey data show AI users save an average 5.4% of their work hours, translating...
The Real AI Bottleneck May Not Be ROI. It May Be Trust.
The article argues that the biggest barrier to AI adoption is not return on investment but trust. While executives now demand ROI metrics, organizations must first establish confidence that AI behaves predictably, is explainable, and can be governed. Trust, as...
The Human First AI Compass for Hotels
Hospital operators are turning to artificial intelligence to offset chronic labor shortages and rising payroll pressure, but many adopt tools without a clear strategy. The article proposes an "AI Compass" that positions AI as a judgment‑amplifier rather than a replacement,...
Google's New Open Source Gemma 4 12B Analyzes Audio, Video — and Runs Entirely Locally on a Typical 16GB Enterprise...
Google released Gemma 4 12B, an 11.95‑billion‑parameter open‑weights model licensed under Apache 2.0. The model runs locally on a standard enterprise laptop with just 16 GB of VRAM, handling raw audio and visual inputs without separate encoders. It offers a 256K token context...
Life Science Update | May 2026
AI tools are accelerating patent drafting for life‑science firms but introduce unlabelled inaccuracies that become permanent USPTO filings. The Federal Circuit affirmed that engineered host cells for gene therapy are patent‑eligible, while the USPTO rolled out the PIER pilot requiring...
NAN124: AI and Trust in Modern Network Automation
In this episode, host Eric Cho talks with Sif Baksh, a veteran network automation engineer and Principal Solution Architect, about the evolution of automation from early scripting in Perl and Tcl to Python becoming the de‑facto language for network engineers....
Reading Between the Lines of Trump’s New Executive Order on AI
President Trump signed an executive order aimed at fostering AI innovation while mitigating cybersecurity risks from frontier AI models. The order creates a voluntary framework that gives the government a 30‑day pre‑release access window to evaluate new models, establishes a...
World Model Taxonomy Clarifies AI’s Overlapping Frontiers
“World model” has become one of the most important and most overloaded terms in AI. Computer vision, robotics, reinforcement learning, and generative AI each claim to be building world models, and each means something quite different. In this essay, we...
Superintelligent Machines May Well Need Us After All
AI’s mathematical capabilities have leapt forward, highlighted by OpenAI’s recent disproof of an 80‑year‑old conjecture and its assistance in cracking a half‑century‑old problem. Yet the article stresses that these breakthroughs gain legitimacy only through human mathematicians who verify and extend...
Open-Source AI Hits Production-Ready Milestone with Ideogram 4.0
🔥 The gap between open-source and frontier AI is shrinking fast. Ideogram 4.0 just became the #1 open-weight text-to-image model on DesignArena, setting a new benchmark for open image generation. What stands out: • Native 2K resolution • Accurate text rendering • Background transparency • Precise...
Microsoft Build 2026: Pushing The Frontier With A More Opinionated AI Playbook
Microsoft Build 2026 delivered a decisive, full‑stack AI playbook, moving from vague guidance to concrete, opinionated solutions across hardware, data, runtime, developer tools, and security. The event showcased the Surface RTX Spark edge‑AI device, the GA of Fabric IQ—a unified...
Reality Defender Integrates Deepfake Detection Into ZeroFox Platform
Reality Defender, a deep‑fake detection specialist, has partnered with ZeroFox to embed its multimodal AI‑generated media detection into ZeroFox’s Discover, Validate, Disrupt workflow. The integration delivers real‑time verdicts on voice, video and image deepfakes via Reality Defender’s RealAPI, letting analysts...
UK Government Urges Companies to Share Data About AI Effects on Workforce
The UK government has asked private firms to provide data on how artificial intelligence is reshaping their workforces. The call, issued by the Department for Business and Trade, seeks information on job displacement, skill upgrades, and productivity gains linked to...
Liferay Introduces Liferay AI Hub?
Liferay announced Liferay AI Hub, a SaaS platform that lets enterprises build, deploy, and manage AI agents through a low‑code studio. The hub leverages Liferay DXP’s security framework, providing granular access controls, audit trails, and compliance certifications such as GDPR,...
Outreach Launches MCP Suite and Agentic Ecosystem Marketplace
Outreach unveiled its MCP Client and MCP Server, creating a unified agentic AI platform for revenue teams. The new Marketplace offers a curated ecosystem of integrations, linking tools such as Seismic, Snowflake, Slack and ZoomInfo to the Outreach stack. External...
Nextworld Launches “Agentic Development” To Bring Governance and Production Readiness to Enterprise AI Software Creation
Nextworld announced the general availability of Agentic Development, a new capability within its platform that extends AI‑assisted application creation from prototype to governed production. The multi‑agent architecture assigns product‑owner, design‑development, and quality‑assurance agents to generate, test, and deploy code under...
Salesforce Launches AI Marketing Agents
Salesforce unveiled its Agentforce suite at the June 3 Connections event, introducing AI‑driven Marketing Agents that assist marketers in building pipelines, generating omnichannel content, and running campaigns. The Content Agent can create emails, SMS, RCS and localized assets from a simple...
SoftBank CEO Masayoshi Son Calls AI the Next 50‑X Tech Revolution After $87 B France Bet
SoftBank chief Masayoshi Son told CNBC in Paris that AI will outpace the dot‑com boom by 10‑to‑50 times, reinforcing his firm’s €75 bn ($87 bn) investment in French AI data centers. The comment fuels optimism for AI‑related equities as Nvidia and other...
Google's Gemini AI Search Overhauls Ads, Sends Advertisers Scrambling
Google announced that its Gemini AI‑powered search will embed "conversational discovery ads" directly in AI answers, sidelining traditional sponsored links. The move, billed as the biggest shift in search advertising in more than 25 years, has left digital marketers uncertain...
Microsoft Fabric Launches Data‑Centric AI Platform, Prioritizing Context Over Model Size
Microsoft announced its Fabric platform, bundling a massive PostgreSQL‑compatible database, GPU‑boosted data warehouse and a semantic layer, to give AI agents enterprise‑level data context. The move targets DevOps teams handling AI‑enabled apps, promising up to 7× performance gains and tighter...
Realtor.com Rolls Out RealAssist™ AI, a Gemini‑Powered Home‑Search Bot
Realtor.com launched RealAssist™ AI, an AI‑first home‑search experience built on Google’s Gemini and Google Cloud, today in a beta for logged‑in users. The tool guides buyers from first question to closing via natural conversation, positioning the company as a pioneer...
Andon Labs Scales Its AI-Run Vending Machine Experiment Into Autonomous AI-Operated Stores and Cafés Within a Year
Andon Labs has expanded its AI‑operated vending machine pilot into fully autonomous retail stores and cafés, eliminating human decision‑makers. The new model relies on a multi‑agent system where a lead "mechanical CEO" coordinates sub‑agents for procurement, customer interaction, and logistics....
Increasing Compute and Tools Obscure True Model Capabilities Pre
Critical point... For example, inference gain (more compute at test time), systems gain (scaffolds, tools, agents), and asset gain (pairing a model with restricted data or hardware) all make it harder to elicit a model's real capabilities before release. Further...
Drug Discovery Demands AI Beyond Language Models
Beyond Language: Why Drug Discovery Needs a Different Kind of AI. Guest editorial by Woody Sherman of PsiThera. https://t.co/HsLyEcZqil
CoreWeave's AI Compute Backlog Swells as Profitability Comes Under Scrutiny
CoreWeave Inc. announced a surge in AI‑compute backlog and a $660 million annual recurring revenue goal, yet its Q4 net loss widened to $76.3 million and operating margin fell to 24%. The mixed signals highlight the tension between rapid growth and near‑term...
Introducing Gemma 4 12B: A Unified, Encoder-Free Multimodal Model
Gemma 4 12B is a new 12‑billion‑parameter multimodal model that eliminates separate vision and audio encoders, feeding raw inputs directly into the language‑model backbone. It delivers benchmark performance close to Google’s larger 26 B Mixture‑of‑Experts model while fitting within a 16 GB VRAM laptop,...
Athenahealth Raises the Bar on Revenue Cycle Management Performance, with 80+ AI-Native Features Designed to Eliminate Manual Revenue Cycle Work
athenahealth unveiled an AI‑native roadmap for its athenaOne platform, embedding more than 80 artificial‑intelligence features to streamline ambulatory revenue‑cycle management. Early deployments have already delivered a 30% lift in recovered payments from coding denials, a 16% drop in insurance‑related denials,...
AI Factories Need Private 5G From Celona
AI‑driven data centers are evolving into high‑density "AI factories" that generate extreme electromagnetic interference, rendering traditional Wi‑Fi unreliable. Celona proposes a private 5G network that delivers roughly 30 dB more noise immunity, maintaining connections down to -90 dBm RSSI. This resilience enables...
Trump’s New AI Order Raises the Stakes in China-US Tech Competition
President Donald Trump issued an executive order on June 2 that repositions artificial‑intelligence development as a core national‑security priority. The order creates a voluntary assessment framework allowing federal agencies early access to frontier AI models and designates the NSA and CISA...
Morgan Stanley Will Soon Open Its Trillion-Dollar Wealth Management Funnel to AI Agents
Morgan Stanley announced it will let external AI agents directly access its stock‑administration platforms, ShareWorks and Equity Edge, for corporate clients. The move targets the firm’s 3,400 workplace‑administration customers and aims to streamline complex stock‑plan management without adding human staff. The...
Still Facing Copyright Lawsuits, AI Music Generator Suno Raises Another $400M
Suno announced a $400 million Series D round that lifts its valuation to $5.4 billion, a steep rise from the $2.45 billion valuation just seven months earlier. The AI music generator faces ongoing lawsuits from Universal Music Group, Sony and GEMA, which allege the...
AI to Find Clogged Drains? Possibilities, Misconceptions and More on AI’s Growth in the Public Sector
Samsara’s senior vice president Tim Nagy says the company’s AI platform is rapidly expanding in the public sector, turning raw sensor data into actionable insights for city services. The technology now processes over 10 million sensor events daily, enabling predictive maintenance that...
South Africa’s AI Skills Gap Is Widening Faster than Universities Can Keep Up
South Africa’s rapid AI adoption is outpacing the country’s education system, creating a widening talent gap. Salesforce executives highlighted that universities struggle to update curricula faster than AI technologies evolve, leaving businesses scrambling for skilled workers. The shortage is most...
AI Has A Constitution Now
The AI safety narrative is shifting from vague slogans like “be helpful” to concrete, written rulebooks that act as a constitution for each model. These constitutions spell out detailed behavioral directives, making safety goals testable and auditable. Leading labs such...
UN Researchers Project AI to Double Data Center Power, Water Consumption by 2030
UN researchers warn that AI‑driven data‑center expansion will double global power and water use by 2030. Consumption is projected to rise from 448 TWh to 945 TWh of electricity and from 4.5 trillion to 9.3 trillion liters of water, while CO₂ emissions could climb...
Google Is Offering Android Developers Cash in Exchange for Code to Train AI
Google is piloting a program that pays Android developers to share the source code of their current and archived Play Store apps. The non‑exclusive licensing deal lets developers retain copyright while monetizing their code for AI training. Google hopes the...
Ditch the Niceties in AI Prompts to Save Energy Use, Say Researchers
UN researchers warn that AI’s energy appetite is soaring and propose a simple fix: trim polite words from prompts. Their study shows that removing filler such as “please” or “thank you” can slash ChatGPT’s power draw by up to 25%,...
Midwestern University Selects D2L to Support AI Innovation in Teaching and Learning
Midwestern University (MWU) announced a partnership with D2L to replace its learning management system with Brightspace, an AI‑enhanced platform. The rollout begins in Fall 2026 and will span the university’s 15 colleges in Phoenix and Chicago. MWU’s 10‑year strategic plan emphasizes...
AI Won't Kill Your Coding Career
In this episode, Brick Thompson and Landon Oaks discuss how AI coding tools have evolved from early, unreliable versions to powerful assistants like Opus 4.7, GPT‑5.5, and Claude Code, enabling developers to generate routine code quickly while still requiring human...
These Two Founders Left Goldman and Meta to Build Voice AI for Markets Everyone Else Overlooked
AethexAI, a voice‑AI startup targeting Africa and the Middle East, closed a $3 million pre‑seed round led by 4DX Ventures. The founders—former Goldman Sachs and Meta employees— built proprietary small‑scale language models and an orchestration layer to overcome latency and dialect...
How ACTICO Is Building AI-Ready Compliance for a Regulated World
ACTICO is rolling out an AI‑ready compliance platform that lets AML officers describe scenarios in plain language and instantly generates deployable, graphical rule sets without IT involvement. The solution addresses three market pressures—AI accountability, data sovereignty, and the rapid pace...
Interview: WSP Experts on How AI Is Being Used to Save Engineers’ Time
WSP’s AI team is embedding machine‑learning tools across its UK and Ireland practice, beginning with Microsoft Copilot to automate routine engineering tasks. The rollout has delivered an average weekly time saving of about 165 minutes per staff member who uses...
‘Human-Made' Is Doing for Music What 'Made In Italy' Did for Fashion.
In 2026 the AI‑music market has settled into a hybrid “human‑made” model where composers create the core track and AI scales the production. This shift mirrors the “Made in Italy” premium label, turning human involvement into a brand differentiator rather...
AI Adoption Surges, but Healthcare Providers Worry About Deskilling
AI use in U.S. healthcare is accelerating, with 74% of doctors and 70% of nurses reporting weekly interactions—a jump from 38% and 46% a year ago. Clinicians rely on AI for tasks such as literature summarization, data analysis, and AI‑driven...
AI-Enabled RCM Needs More than Just Good Tech
Omega Healthcare CEO Anurag Mehta says AI‑enabled revenue cycle management (RCM) must go beyond technology, emphasizing data quality, population health insights, and provider‑specific workflows. His firm partners with hospitals to customize AI tools that reflect each organization’s patient mix and...
C.H. Robinson Expands AI Portfolio with Lean AI Engineer for Continuous Supply Chain Optimization
C.H. Robinson launched Lean AI Engineer, a closed‑loop AI service that continuously monitors and improves global 4PL shipments. Paired with the existing Lean AI Planner, the system now autonomously handles 92% of shipments across trucking, ocean, air and rail, learning...
AI Factories Go Utility-Scale as Siemens Debuts 100 MW IT Design
Siemens unveiled a 100 MW AI‑factory reference architecture, developed with Nvidia, Fluence and Vertiv, that bundles electrical distribution, battery energy storage and liquid cooling into a single, utility‑scale blueprint. The design supports 136 MW of total facility power, offering a repeatable, factory‑tested...