Today's AI Pulse
Salesforce to acquire AI‑agent platform Fin for $3.6B
Salesforce announced a $3.6 billion purchase of Fin, the AI‑agent platform formerly known as Intercom, to strengthen its Agentforce suite. The deal comes as Salesforce reported strong Q1 results and a $50 billion buyback program. Fin’s technology enables automated customer support across chat, email, WhatsApp, phone and Slack.
Also developing:
By the numbers: Waymo acquires Apple’s Arizona Proving Ground for $220M
The Regulation Dilemma: Balancing AI Innovation with European Complexity
European firms are racing to adopt generative AI while the EU AI Act tightens rules for high‑risk systems, especially with Article 72 now in force. Executives are urged to embed AI directly into secure, governed workflows rather than layering on disparate tools. Proactive governance—assigning a clear owner and using centralized platforms—can turn compliance into a competitive advantage. ServiceNow positions its AI control‑tower as a solution to achieve this governed acceleration.
AI‑Driven Benefits Platform Raises $30M to Transform HR
Origin Secures $30 Million Series A+ to Scale AI‑Powered Global Benefits Platform Origin’s financing highlights a growing consensus that AI can unlock efficiency in traditionally manual HR functions. By centralising disparate benefits data, the platform not only cuts costs but also...
How to Spot Artificial Intelligence Recruiters Who Target Candidates From LinkedIn
Research firm Gartner predicts that by 2028 one in four job candidates worldwide will be fabricated, fueling a surge in AI‑generated recruiter outreach. Executives are receiving polished, generic emails that often originate from Gmail accounts and contain vague role descriptions,...
Spacebring Launches Lem AI Agent For Coworking Space And Flex Office Operations
Spacebring, a platform serving more than 500 coworking locations worldwide, has launched Lem AI agent to automate routine administrative tasks. The AI assistant handles support ticket triage, retrieves operational data via natural‑language queries, and drafts community communications. Lem AI is...
Jump Launches AI Associate
Jump has unveiled AI Associate, an agentic AI layer that lets financial‑services advisors act on insights in real time. The tool integrates with CRM, email, and planning systems, enabling users to pull data, update records, draft communications, and schedule meetings...
Watch James Manyika Talk AI and Creativity with LL COOL J.
James Manyika, Google’s senior vice president for research, labs, technology and society, sits down with hip‑hop legend LL Cool J for the latest Dialogues on Technology and Society episode. The pair discuss the evolution of creative tools, from early drum machines to...
Collectly Acquires AI Automation Startup Pledge Health to Streamline Pre-Service Workflows
Collectly, an AI‑driven patient financial platform that has processed over $1 billion for more than 3,000 facilities, announced the acquisition of Y Combinator‑backed Pledge Health. The deal adds upstream pre‑service automation—coverage verification, cost estimates, and prior‑auth readiness—to Collectly’s existing post‑visit billing suite....
Opt-Out Is Not Consent
GitHub announced that, starting April 24, it will automatically use interaction data from Copilot Free, Pro, and Pro+ individual users to train its AI models unless users manually opt out. The data includes code, file names, repository structure, cursor context,...
China Unicom Bets on AI Cloud Services to Break Telecoms Revenue Ceiling
China Unicom is accelerating AI deployment across consumer, enterprise, and home services, targeting 2026 as a launch point for new revenue streams beyond traditional voice and data. Its flagship AI Cloud Phone, built with Huawei, already has 18 million users and...
Finance Ops Drives 45% AI Agent Market Surge
the ai agents market grew from $8.29 billion in 2025 to $12.06 billion in 2026. that's 45.5% year over year. and finance operations is the number one entry point. 33% of enterprise leaders are piloting agentic ai in finance ops first, ahead...
A Hybrid CNN–Transformer Network to EnhanceSolar Magnetogram Resolution for Flare PredictiveAnalytics
Researchers introduced MagRes‑Net, a hybrid convolutional neural network and transformer architecture that upscales low‑resolution SOHO/MDI magnetograms to match the detail of SDO/HMI observations. Trained on co‑aligned MDI‑HMI image pairs with physics‑aware constraints, the model restores fine magnetic structures while conserving...
I Teach at Harvard and Encourage My Students to Use AI on Every Assignment. They Just Have to Follow My...
A Harvard professor has made AI a mandatory tool for every student assignment, but only as a research aid and editorial assistant. Students must first formulate their arguments independently before turning to AI for polishing and gap‑filling. The professor frames...
The Rise of the AI Super Agent: How HIX.AI 2.0 Is Redefining the All-in-One AI Workspace
HIX.AI 2.0 launches an all‑in‑one AI workspace that replaces fragmented tool stacks with a Coordinator‑driven, autonomous task planner. The platform combines research, data structuring, multi‑format content creation, and lightweight coding into a single, skill‑based architecture. Users can initiate complex workflows—such...
Security Researchers Sound the Alarm on Vulnerabilities in AI-Generated Code
Georgia Tech’s Vibe Security Radar identified 35 new AI‑generated code vulnerabilities in March 2026, raising the quarterly total to 74 confirmed CVEs linked to AI coding tools. The project tracks roughly 50 AI‑assisted development platforms, with Anthropic’s Claude Code accounting for...
Quail Digital and Audivi Join Forces to Bring Cost-Free Voice AI to Drive-Thru Restaurant Ordering
Quail Digital and Audivi have announced a global partnership delivering a free Voice AI hardware box for quick‑service restaurants. The solution merges Quail’s high‑clarity audio equipment with Audivi’s real‑time AI ordering software, aiming to cut drive‑thru errors and speed service....
Build AI-First Teams: Small, Senior, Product‑Driven
Zero-Based Engineering: Rethinking Your Team for the AI Era My business partner Shiyan Koh recently shared a post that I found illuminating. Most founders are asking the wrong question about their engineering teams. Instead of "who can I cut?" they...
Scoop: Meta Gets Ready to Launch Two New Ray-Ban AI Glasses
Meta and its hardware partner EssilorLuxottica are preparing to launch two new Ray‑Ban AI glasses, as indicated by recent FCC filings that list the devices as production‑ready units. The filings suggest the models are close to market, echoing the rapid...
The "Negative Space" Prompting Technique
A robotics startup in the Pacific Northwest struggled with AI‑generated documentation that hallucinated features and added marketing fluff, delaying product launch. The engineering team’s standard prompts produced inaccurate safety manuals and API docs, requiring extensive manual editing. The author introduced...
Lucid Software Advances MCP Server Offering and Lucid AI Capabilities, Debuts Process Agent to Boost Speed and Clarity Across Teams
Lucid Software unveiled a suite of AI‑enhanced collaboration tools, including an upgraded Model Context Protocol (MCP) server that can generate editable diagrams from natural‑language prompts. The company added voice‑to‑text prompting, intelligent layout features, and an Edit Document API slated for...
Accenture Introduces Cyber.AI Platform Powered by Anthropic Claude
Accenture has launched Cyber.AI, an AI‑driven cybersecurity platform built with Anthropic’s Claude model. The solution combines autonomous agents with Claude’s reasoning engine and includes Agent Shield for real‑time governance of AI agents. In Accenture’s own environment, the platform secured 1,600...
Copilot to Train on GitHub, Security Agents Comes Free(ish) to 365 E5
Microsoft announced that GitHub Copilot will continue training on publicly available GitHub code, but enterprise customers are excluded from contributing data to the model. At the same time, Microsoft 365 E5 subscribers will receive Microsoft’s security agents enabled by default...
AI Widens Intelligence Gap Through Engagement, Not Ability
🔥I’ve been thinking about the popular post by @DrSuneelDhand that AI will make smart people smarter and dumb people dumber. AI doesn’t move the center of intelligence so much as it stretches how we use it. The average stays where...
Revenium Introduces AI Outcomes to Measure ROI at the Agentic Workflow Level
Revenium launched AI Outcomes, a solution that ties every AI agent execution to a specific business result and calculates ROI at the workflow level. The platform adds an outcome identifier to each step, capturing execution status and the downstream business...
ByteDance Rolls Out Seedance 2.0, as OpenAI Scraps Sora
ByteDance unveiled Seedance 2.0, a text‑to‑video model that produces near‑Hollywood‑quality clips, prompting immediate backlash from major studios over potential copyright infringement. The company paused its global rollout while expanding the service in Africa, Latin America, the Middle East, and Southeast...
Automate Your Inbox, Gain a Day per Hour
When I'm not working, I'm building. And I'm an inbox zero, so I built a system where Claude checks my inbox every 5 minutes and pings me on Slack with what isn't worth my attention and should be archived. Every hour I...
Intercom's New Post-Trained Fin Apex 1.0 Beats GPT-5.4 and Claude Sonnet 4.6 at Customer Service Resolutions
Intercom unveiled Fin Apex 1.0, a purpose‑built AI model that powers its Fin customer‑service agent handling over two million weekly conversations. Benchmarks show a 73.1% resolution rate, edging out OpenAI’s GPT‑5.4 and Anthropic’s Claude models by roughly two percentage points....
Comparing AI Memory Benchmarks Reveals Key Design Trade‑offs
[Opus 4.6 Research] "Research the top memory benchmarks for AI systems, look at the top performing memory systems, and to the extent you can look at how they approach memory (looking at code if they are open source), note similarities...
Genetic Map of Brain Aging Meets MIND Diet: 2.5‑Year Slowdown Unveiled
An international team led by Zhejiang University and USC researchers has released the first region‑by‑region genetic map of brain aging, while a separate study of 1,647 adults finds the MIND diet can delay brain aging by roughly 2.5 years. The...
The Path To AI Maturity: What Leaders Should Consider In The Coming Year
AI has reached a tipping point in the legal sector, moving from experimental pilots to a baseline expectation for many firms. Litify’s 2025 State of AI in Legal Report shows rapid adoption but uneven maturity across organizations. A March 31 webinar...
Argonne Researchers Develop AI System to Enhance Electric Grid Efficiency and Reliability
Argonne National Laboratory unveiled GridMind, an agentic AI co‑pilot that lets power‑grid operators run scheduling, weather‑impact and reliability simulations through natural‑language conversation. The system uses a multi‑agent architecture coordinated by large language models such as GPT‑5, GPT‑4o and Claude 4 Sonnet,...
Clients Demand AI Speed, yet Crave Hand‑crafted Look
Spoke to someone recently doing big commercial client work saying clients insist on AI speed and flexibility and yet insist on output looking believably like manual 3d generations so they don’t get backlash over it It’s all a bit silly and...
Origin Secures $30 Million Series A+ to Scale AI‑Powered Global Benefits Platform
Origin raised $30 million in a Series A+ round led by Notion Capital, bringing total funding to over $50 million. The AI‑powered platform promises a single source of truth for multinational benefits, targeting cost‑pressured HR, finance and risk teams.
Colleges Turn to Oral Defense Exams to Thwart AI-Generated Work
Cornell, the University of Pennsylvania and New York University are rolling out in‑person oral defense exams to verify student learning as generative AI tools proliferate. Faculty say the move is a direct response to flawless AI‑written essays that mask gaps...
Jabil Beats Q2 2026 Guidance, Raises Full-Year Revenue Outlook to $34B
Jabil reported Q2 2026 net revenue of $8.3 billion, topping guidance by $500 million, and lifted its full‑year revenue forecast to about $34 billion. CFO Gregory Hebard highlighted robust growth in Intelligent Infrastructure and AI‑related services while reaffirming a capital‑allocation plan that favors...
Banks Must Measure AI Impact, Not Training Completion
Redesign Work, Realize AI: How Banks Must Measure and Reorganize to Turn AI from Experiments into Durable ROI Stop measuring training completions. Start measuring: cycle time, cost-to-serve, throughput, risk outcomes. Three-dimension framework: capability adoption (role-specific literacy, agent management) + workflow transformation...
AI Agents: Future Economic Players or Mere Tools?
Are AI agents going to be real economic actors — or just tools? We break down the competing visions. https://t.co/9SSDq5hVSn
NYT’s AI‑Generated Modern Love Column Sparks Data‑Governance Debate
The New York Times published a Modern Love essay that AI‑detection tools flagged as more than 60% generated by artificial intelligence. The incident has sparked a clash between journalists, AI researchers and editors over data‑governance, bias and disclosure standards in newsrooms.
Claude Automates Stack Setup, Making Deployment Effortless
Struggling to get a production stack together? Just get Claude to do it. Need to pay for it? Just get Claude to do it. Bit by bit autonomy is coming.
AI Governance Rises Globally—Readiness No Longer Optional
48 hours this week: US: constitutional protections for AI outputs. Europe: enforceable compliance requirements. Scotland: five-year national AI strategy. Three continents. One conversation. The governance infrastructure is being built. Your company's readiness isn't a factor.
Multiply Raises $9.5M, Focused on B2B Self-Learning Ads
Multiply, an AI‑native media agency for B2B firms, emerged from stealth with a $9.5 million Series A led by Mayfield. Its self‑learning advertising platform links Google Search and LinkedIn campaigns to sales call recordings, CRM data, and performance metrics, automatically generating and...
China's S1 Humanoid Robot Operates Retail Kiosks
Meet S1: The Humanoid #Robot Running Retail Kiosks in China by @CyberRobooo #Robotics #Engineering #ArtificialIntelligence #Innovation #Technology https://t.co/cm3OJSAIR2
Can AI Master Empathy: State or Language?
Can AI "learn" empathy? Is empathy an inner social state or a set of linguistic strategies? https://t.co/REyUpNkS7H
AI Agents Trigger Payment Wars Among Tech Giants
Stripe. Coinbase. Visa. Everyone is building for AI agents. Now LIVE on Unchained, we unpack the battle for the payment layer of the agent economy. https://t.co/meE1SvxnF0
AI: Handy Shortcut, But Overuse Breeds Hidden Complexity
AI is a shortcut. So it’s useful. But it’s lazy. So it speeds execution. But it hides complexity. So you want to use it. But definitely not overuse it. So the user of AI often loves it. But the reader of AI often hates it.
AI Becomes Society’s New Horizontal Infrastructure
This is so well put. AI will be as ubiquitous in future societal advances and catastrophes as electricity or books. It's just a new, horizontal part of society's infrastructure.
Right Data & Context Transform Zero‑Shot AI
Very much agree with this Cowork zero shot is meh When connected to the right data and fed with deep, detailed context: it’s remarkable And once set up, the MCP and Skills architecture make it quite usable
Twitter Leverages AI for Real‑Time SEO, Threads Lags
The difference between Twitter and Threads right now: Threads: AI can’t do SEO research for you it doesn’t have access to real time data. Twitter: I just hooked up Claude to SEMrush MCP and setup 5 skills to handle all my...
Bots Grow Eight Times Faster Than Humans Online
Already, wow -> AI and bots have officially taken over the internet, report finds "HUMAN Security’s State of AI Traffic report found that automated traffic grew eight times faster than human traffic year-over-year." https://t.co/NFmGuouIgV
CLAUDE.md Can Quietly Mislead Your Code Agent
Wrote up how we built a Claude Code skill that audits project docs against source code. The most interesting finding? Your CLAUDE.md can silently misdirect your agent.
Singularity Shatters All Existing Business Models and Economics
The Singularity is a firewall: beyond it, every business model and economic theory you trust gets vaporized.