Today's AI Pulse

Salesforce to acquire AI customer‑service platform Fin for $3.6B
Salesforce announced a $3.6 billion acquisition of Fin, an AI‑driven customer‑service platform, slated to close in fiscal Q4 2027. Fin’s AI agents handle chat, email, WhatsApp, text, phone and Slack, resolving roughly 76% of tickets without human intervention. The deal adds over 30,000 enterprise customers and $400 million in annual recurring revenue.
Also developing:
By the numbers: Waymo acquires Apple’s Arizona Proving Ground for $220M

Study Finds A Third of New Websites Are AI-Generated
Researchers analyzing Internet Archive data found that roughly 35% of websites launched between August 2022 and May 2025 were AI‑generated or AI‑assisted, up from virtually none before ChatGPT’s debut. Using the Pangram v3 detection tool, the team examined semantic diversity, sentiment and factual accuracy across millions of archived pages. Their results show AI content makes the web more upbeat and less semantically dense, but it does not increase verifiable misinformation. The authors plan to turn the study into an ongoing monitoring system to track AI’s evolving impact on the digital landscape.
Technology Does Not Belong to the Technologists
Sam Altman released a set of OpenAI principles asserting that artificial intelligence will dwarf the impact of steam engines, electricity, and even modern computing. The author of the article rebuts this claim as hubristic, pointing to history’s unpredictable consequences when...

The AI Safety Movement Needs Normies
The AI safety community’s recent focus on existential risk has left it isolated from the broader public, a gap highlighted by a March protest outside Anthropic and a Molotov‑cocktail attack on Sam Altman’s San Francisco home in April. While activists demand...
Energy Capacity Becomes AI Industry’s New Status Symbol
AI companies are starting to compete on energy capacity. Claims about access to massive amounts of power are becoming a way to signal scale and strength, even when the numbers are hard to verify. The shift is revealing. In AI, computing and...
University of Southampton and UCB to Develop Digital Antibodies with AI
The University of Southampton and biopharma firm UCB have launched the Digital Antibody Research Collaboration (DARC), a three‑year AI‑driven platform for in‑silico therapeutic antibody design. DARC aims to compress the typical ten‑to‑twelve‑year development timeline by automating molecule modeling, testing and...
DeepSeek V4 Arrives With Near State-of-the-Art Intelligence At 1/6th the Cost
DeepSeek, the Chinese AI startup spun out of High‑Flyer Capital, launched DeepSeek‑V4, a 1.6‑trillion‑parameter Mixture‑of‑Experts model, under an MIT open‑source license. The model matches or exceeds leading closed‑source systems on several benchmarks while costing roughly one‑sixth of GPT‑5.5 and Claude...

I Tried Copilot’s New Tools for Word, Excel, and Powerpoint, and I’m Not Sure I Will Again
Microsoft has made its AI‑driven Copilot agentic features generally available across Word, Excel, and PowerPoint for all paid Microsoft 365 subscribers. The tools now draft whole documents, build spreadsheets and charts, and generate slide decks based on natural‑language prompts, moving beyond...

Meet Your Agentic Contact Center
Microsoft unveiled Dynamics 365 Contact Center, a unified AI platform that combines three purpose‑built agents—Customer Assist, Quality Assurance, and Service Operations—into a single Copilot‑driven control plane. The Customer Assist Agent now supports real‑time voice AI with low‑latency, multi‑language handling, while...

Fama Technologies Granted U.S. Patent on Foundational AI-Powered Social Media Screening Technology
Fama Technologies announced that the USPTO has allowed all 22 claims in its patent covering the foundational AI‑powered social‑media screening system, with a priority date of November 30, 2015. The patent confirms Fama as the original inventor of the technology that extracts...

GenAI Citations, Explained
Generative AI tools such as ChatGPT, Claude, Gemini and others now attach source links—called citations—to their answers. The article reveals that Google powers citations for ChatGPT, Gemini, AI Mode and Grok, while Brave supplies citations for Claude and Perplexity, making high...

Will Gravity Pull Down the AI Bubble?
Dean Baker argues that the current surge in AI‑related equities mirrors the late‑1990s dot‑com frenzy, with sky‑high valuations, fierce global competition, and uncertain profit paths. He points to a widening gap between market hype and underlying fundamentals, noting that rising...
Make Your Site Visible to AI with Structured Data
AI agents don't "read" your homepage. They scan your code. If they can't pull clean, structured info in seconds, you're invisible. 5 things you need to help AI find you: 1. Schema markup 2. Clear copy, no fluff, just say what you mean 3....
Pharma Meets AI Conference 2026: AI Governance Shifts to Embedded Practice
At the Pharma Meets AI conference in Barcelona, industry leaders announced a pivot from standalone AI policy frameworks to embedded governance models. Novartis presented an “integrated assurance” approach that fuses risk management, compliance, and internal controls directly into business processes....

Ineffable Intelligence Launches with Record-Breaking $1.1B Seed Round
Ineffable Intelligence emerged from stealth with a record‑breaking $1.1 billion seed round, valuing the startup at $5.1 billion. Founded by DeepMind veteran David Silver, the company aims to build a “superlearner” that discovers knowledge autonomously using reinforcement learning. The fundraising was co‑led...
World Models: Key to AI’s Physical World Integration
AI companies want to build systems that understand the external world. If they succeed, they may overcome limitations of LLMs and help AI enter physical environments. That's why we named world models one of the 10 Things That Matter in...
AI Drives S&P Tech Profits Toward Semiconductors
Macro: AI shifts S&P tech profits to semiconductors. Key: semis ~42% of IT cap, large earnings upgrades; software faces obsolescence. Trade: overweight semiconductors vs software. — Viktor Kopylov, PhD, CFA More insights: t.me/si14Kopylov

AI Is Creating a Shadow Version of You
The post argues that generative AI systems create a “shadow” version of each user by learning behavioral patterns rather than explicit preferences. Through repeated prompts, edits, and timing cues, the model builds a compressed outline of how the user writes...

‘Digital Me’ Is Turning Human Capability Into Corporate Assets. HR Must Push Back
The article warns that the emerging "Digital Me"—AI‑driven replicas of an employee’s knowledge, judgment and decision‑making—is turning individual capability into a reusable corporate asset. As these digital twins can generate value long after a worker leaves, traditional contracts that treat...
Ambient AI Tools Improve the Patient-Clinician Connection
Ambient AI, championed by Abridge, captures entire patient visits and drafts clinical notes in real time, allowing clinicians to stay fully present during encounters. Matt Troup, Abridge’s clinical strategy principal, says the technology eases cognitive burden and helps patients feel...

AI All the Way Down
A new arXiv paper simulates 144 synthetic astrophysicists to test large language model (LLM) assistance across 2,592 research tasks. Using Qwen3:8B and DeepSeek‑R1, the study finds AI improves writing‑heavy activities but harms quantitative derivations, often producing confident yet wildly incorrect...
Partnership for Clinical Decision Support Moves AI to Next Level
AI-driven clinical decision support has taken a leap forward as Abridge announced a partnership with the New England Journal of Medicine and JAMA. The collaboration will embed peer-reviewed articles and guideline evidence directly into Abridge’s platform, linking them with real-time...

Collaborative Develops AI Vendor Disclosure Framework
The Health AI Partnership (HAIP) has published an AI Vendor Disclosure Framework in NEJM AI, offering a free, publicly‑available tool for health‑system leaders evaluating AI solutions. The framework outlines five core domains—system capabilities, performance, data stewardship, integration requirements, and lifecycle management—to...
Nvidia’s ‘AI Insurance Policy’ Balances Immediate and Future AI Approaches
Nvidia is diversifying beyond its cloud‑centric AI revenue by promoting on‑device LLMs, real‑time world‑model AI, and GPU‑based quantum simulation. The company highlights ChatRTX for RTX‑30 GPUs, expands AI‑driven digital twins and robotics, and rolls out CUDA‑Q and cuQuantum for quantum...
DeepSeek V4 Handles Far Longer Prompts Efficiently
Chinese AI firm DeepSeek released a preview of V4, its long-awaited new flagship model. Notably, the model can process much longer prompts than its last generation, thanks to a new design that handles large amounts of text more efficiently.

AI Physics Models Cut Design Time From Weeks to Minutes
AI models trained on physics simulations are speeding up design and development across engineering. @generalmotors’s in-house large physics model returns results in a matter of minutes for a process that used to take weeks. https://spectrum.ieee.org/large-physics-models-design-engineering

From Casebook to Copilot: Bridging Law's AI Readiness Gap
Law schools are falling behind on artificial‑intelligence training, creating a widening gap between academic curricula and the expectations of modern law firms. A recent survey shows fewer than 10% of law‑school courses now cover generative AI, while 65% of firm...

AI's Trust Crisis: Leaders Must Rebrand Now
Fear is rising. Trust is falling. And the AI industry has no one to blame but itself. AI leaders like Dario, Sam, Zuck, and Mustafa can do something about it. https://briansolis.substack.com/p/ai-has-a-branding-problem-ai-native

I Tested ChatGPT Images 2.0 Vs. Gemini Nano Banana to See Which Is Better - This Model Wins
OpenAI unveiled ChatGPT Images 2.0, a next‑generation image generator that adds real‑data context and embedded text. In a head‑to‑head test against Google Gemini’s Nano Banana, ChatGPT Images 2.0 achieved a 97% normalized score (150 points) while Nano Banana fell to 85% (131 points). The...

Verda Bets $117M on AI Cloud as Workloads Fragment Across Platforms
Finland‑based AI infrastructure firm Verda secured $117 million in equity and debt financing to accelerate platform development and expand into the US, Europe and Asia. The funding backs Verda’s niche strategy of serving high‑intensity AI training and inference workloads that demand...
Turn Sent Emails Into AI-Powered Knowledge Base
I was working with a client this week on connecting their email to Notion, and they mentioned something I hadn't quite considered. They spend two hours every morning answering customer support emails. Same questions, over and over. How does delivery work?...
AI Translation Is Possible, but Not Always Advisable
Just because you *can* use AI to translate and scale your website into many different languages does not mean you *should.*

OpenAI and Microsoft Rewrite Their Deal: No More Exclusivity, No More AGI Clause
OpenAI and Microsoft have renegotiated their partnership, ending Microsoft’s exclusive Azure rights and removing the controversial AGI clause. OpenAI can now distribute its models on any cloud, though Azure remains the launch platform and primary partner. The financial terms shift...

Scaling AI Often Means Scaling Chaos, Not Value
Everyone thinks they’re scaling AI. Most are just scaling chaos. What I’m seeing, including in conversations with @collibra, is a dangerous illusion → faster experimentation → more pilots → zero control The gap between AI ambition and actual business value is widening. Here’s what’s really going on… #CollibraPartner
AI Reveals Life's Simplicity After Decades of Complexity
AI tools are showing everyone how simple life can be. We've been running around in circles for decades.

While 80% of AI Projects Fail, One UK Planning Consultancy Has Already Made It Pay
Planning By Design, a UK planning consultancy, has turned AI into a profit centre by embedding a custom assistant, Aisa, across customer engagement, project management, sales, and content production. While RAND estimates 80% of AI projects flop, the firm’s problem‑first,...
Altman's AI Hype Meets Historic Reality and Banality
Altman: "AI will dwarf what people could do with steam engines or electricity." Uh, history would like a word, Sam. Then what follows in his principles is the kind of sophomoric banality only an LLM could produce. Our principles https://t.co/Xxcm5GsU1N
AI Real-Time Beach Volleyball Analysis: Game Boost or Soul Shift
#AI Referees the Sand: Real-Time Beach Volleyball Analysis — Enhancing the Game or Changing Its Soul? by @measure_plan #ArtificialIntelligence #MachineLearning #ML https://t.co/xa7XdYWHHU

Talk to Me, Arthur.
Sugar Capital’s founder recounts rebuilding the firm’s internal workflow as Project C6, a custom AI‑powered platform built in 5.5 weeks. The centerpiece, an AI named Arthur, scans founder decks and returns a 90‑second critique, score, and polished “steel‑man” version. The...

4,800 Disney/ESPN Staff Used AI in Nine Days
“Screenshots of the dashboard viewed by Business Insider show AI usage by about 4,800 product and tech employees across Disney Entertainment and ESPN in a nine-workday span during mid-April.” 👀 https://t.co/dJy7Idxgt2
AI at HMRC Becomes Tax‑penalty Enforcer, Not Helper
> you work at HMRC > you're told you're going to get an AI to improve your working life > get excited > AI arrives, it's Copilot > realise you work at HMRC > go about ruining someone else's day https://t.co/bmiN0DxLJz

Move 37, Ten Years On: Partnering with Ineffable Intelligence
David Silver, the architect behind AlphaGo, AlphaZero, MuZero and AlphaProof, is launching Ineffable Intelligence, a startup focused on AI that learns from its own experience rather than human data. The company is raising a seed round, co‑led by investors who...

Why Elon Musk and Sam Altman Are Fighting over OpenAI
Elon Musk, co‑founder of OpenAI, has filed a lawsuit demanding more than $130 billion in damages, accusing the company and CEO Sam Altman of breaching a non‑compete agreement and misusing his AI‑related intellectual property. The case, filed in a U.S. federal...

AI Hype Ignores Physical Limits and Real Constraints
A brave new world of AI efficiency & magic is a lovely bedtime story @CRudinschi It's the fake techno God complex. That's not how reality works. Reality is about physical limits not make-believe pushed by corporations and politicians. https://t.co/GaanGor64s
Disney's AI Trains on Competitors' Data, Fair Use Questioned
what data were these AI models trained on? answer: data that includes data from disney’s competitors. which means disney is using their competitors data to compete against them in the market. which is fine if fair use applies, which will be...

6 AI Prompts to Standardize Your Solopreneur Business Before You Automate Messy Workflows
The post urges solopreneurs to lock down a precise offer, buyer profile, outcome, and price before layering AI automation on messy workflows. It highlights that tools like Zapier or Claude merely amplify whatever process you feed them, often resulting in...

Google's Gemini AI Likely to Become “Daily Brief”
Gemini's upcoming daily assistant feature could get a logical new name Moving away from the "Your day" moniker, Google appears to be settling on "Daily brief" for its proactive AI hub. ✅ Details - https://t.co/K4qNdw0PZd https://t.co/XsK3DxrPIF
Master AI Agent Management and KPI Realignment
Are You Ready to Navigate Vibe Coding Realities, Manage AI Agents, and Revisit KPIs? https://t.co/KLl48IhHAt via @star_cio #AI #CIO

How Schools Introduce AI To Young Students
AI education is gaining traction in U.S. K‑12 and higher‑education settings, though a cohesive national strategy remains absent. State actions, such as Ohio’s 2025 AI policy deadline, have spurred local rules, with 31 % of high schools adopting policies by August 2025,...

The Data You Give Away Is the Advantage You Lose
The AEC sector is rapidly adopting AI, but firms are inadvertently feeding proprietary project data into platforms that train models usable by competitors. Most enterprise contracts grant vendors rights to anonymized data, a loophole that can expose sensitive win/loss histories...

How AI Is Changing The College Application Experience
Artificial intelligence is reshaping college admissions as students and teachers increasingly use AI to draft essays and recommendation letters, prompting many schools to de‑emphasize or ban the traditional essay. Detection tools flagged 42% of a sample of 1,000 essays, while...