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
Google Tweaks Chrome AI Privacy Wording, Insists Processing Stays On-Device
Google revised the wording in Chrome’s Settings UI that described its on‑device AI, removing the phrase that assured data never leaves the device. The change sparked speculation that Chrome might start routing AI interactions to Google servers, but Google clarified the Gemini Nano model continues to run entirely on the user’s machine. The update coincided with the rollout of the Prompt API, which lets web pages invoke the local model, shifting any data handling to the site’s privacy policy. Users can still disable and uninstall the 4 GB Nano model through Chrome settings.
Pennsylvania Sues over Chatbot Therapy Misuse
People use chatbots as therapists. Describe symptoms, ask about medication. We've been doing it for years. Pennsylvania just put it in a lawsuit.
Only Data Centers with Silicon Truly Count
A new framework I built while researching the global AI buildout. The Empty Shell Test. Simple rule: A data centre announcement only counts when there's silicon committed to go inside it. No chips. No factory. Brazil announced 8 GW. They've delivered ~250 chips. Norway announced 0.8...
Apple Dominated the 2026 Chip War. Google’s Partners Are Left to Buy Scraps
Apple secured over half of TSMC’s 2 nm capacity for 2026 and prepaid multi‑year HBM memory using its $123 B cash reserve, cementing a vertical hardware advantage. Google and partners are left fighting for limited 3 nm wafers and face spot memory prices...
Jensen Huang Just Made a Major Announcement. Here's What It Means for Nvidia Investors.
Nvidia announced Ising, the world’s first open‑source quantum AI model family, extending its AI leadership beyond GPUs. The move positions the company as the software architect for future quantum computers, complementing its $5.1 trillion market cap driven by the AI boom....
SAP Draws a Perimeter Around Agentic AI and What That Means for the Rest of US
SAP released version 4 of its API Policy in April 2026, explicitly prohibiting semi‑autonomous or generative AI systems from calling undocumented or private SAP APIs. The rule limits agentic AI to SAP‑published endpoints and to pathways that SAP has officially endorsed,...
The Supervision Gap
The article highlights a growing supervision gap as firms rush AI into client‑facing workflows. While AI can reliably detect issues such as negative sentiment, translating that signal into the right action demands human context and clear accountability. Real‑world examples from...
Anthropic Partners with SpaceX to Expand AI Capacity
Anthropic announced a computing partnership with SpaceX that will give the company access to 300 MW of AI‑focused compute power at SpaceX’s Colossus 1 facility in Memphis, which houses more than 220,000 Nvidia GPUs. The deal is part of a broader strategy...
Power Shortages Threaten Kazakhstan’s $1.9 Billion Data Center Push
Kazakhstan has signed a memorandum to build a Tier IV data center worth up to $1.5 billion, paired with a $400 million, 250 MW gas‑fired power plant. The initiative is a cornerstone of President Kassym‑Jomart Tokayev’s plan to position the country as a regional...
Close the Loop: Real Traffic Validates AI Experiments
You need to close the loop. Most of you aren't. Everyone's using AI to spin up 50 landing pages, 150 ad variations, entire feature sets in a weekend. It feels like you're testing everything. But you're not. You're just guessing at scale. If you...
Anthropic Just Weaponized the Palantir Model. The Entire Services Industry Is Now in the Crosshairs.
Anthropic announced a $1.5 billion joint venture with Blackstone, Hellman & Friedman and Goldman Sachs to launch an AI‑native enterprise services firm. The venture, backed by additional investors such as General Atlantic and GIC, will embed Anthropic engineers and models directly into mid‑size companies,...
Async Infrastructure Paved Path to AI Agent Era
https://t.co/EAT6n7xgeL lets developers add AI agents to their products with a simple SDK, handling execution, long-running workflows, and reliability so they don't have to. Over 90% of their usage now comes from agent workflows, and they recently announced their $16M...
Wells Fargo Rolls Out $1 B AI Advisor Platform to Modernize Wealth Management
Wells Fargo has launched Advisor Gateway, a $1 billion AI‑driven desktop platform that gives its financial advisors one‑click access to more than 200 tools, including BlackRock’s Aladdin Wealth with generative AI. The rollout aims to streamline client‑facing workflows and keep the...
Sony CEO Unveils AI‑Driven Game Development Strategy, Citing $700M Revenue Boost
Sony Interactive Entertainment president Hideaki Nishino announced an AI‑centric roadmap that promises to slash animation cycles, expand content volume and generate more than $700 million in incremental revenue, signaling a shift that will ripple through game marketing and creative pipelines.
Magna International Deploys AI Across $42 B Auto Parts Empire to Boost Factory Automation
Magna International is rolling out artificial‑intelligence tools across its 330 plants, aiming to improve product quality, equipment uptime, safety and energy efficiency. The move, described by SVP of R&D Sharath Reddy as an "amplifier" for supply‑chain resilience, marks the biggest...
Goldman Sachs CIO Prioritizes Idea‑to‑Production Speed Over AI Usage Tracking
Goldman Sachs chief information officer Marco Argenti says the firm does not track each employee’s AI tool usage. Instead, he gauges productivity by the velocity at which his 12,000 engineers turn ideas into production‑ready software, a shift that challenges the...
PubMatic Expands Agentic AI to Indirect Programmatic Deals, Sparking 80% YoY Revenue Surge
PubMatic announced that its AgenticAI platform will now handle end‑to‑end programmatic deals beyond direct publisher contracts. The move helped the company post an 80% year‑over‑year rise in its AI‑driven revenue segment, lifting total Q1 revenue to $62.6 million. Executives say the...
Continuum AI Unveils OrcaRouter, Open-Source LLM Router with Zero Markup
Continuum AI announced today the launch of OrcaRouter and OrcaRouter Lite, an open‑source, MIT‑licensed inference layer that routes traffic across more than 200 frontier and open‑source language models with zero markup on bring‑your‑own‑key traffic. The move challenges incumbent routers that...
KPMG India Teams Up with CleverTap to Boost AI‑Driven Customer Engagement Consulting
KPMG India has entered a strategic alliance with AI‑first martech platform CleverTap, merging the firm’s transformation consulting with CleverTap’s real‑time analytics and AI‑driven engagement suite. The partnership targets reduced churn, higher lifetime value and compliance‑ready customer ecosystems for corporate clients...
Spacelift Hires Two Senior VPs to Accelerate AI‑Native Infrastructure Automation
Spacelift announced the appointment of John Henry Archer as senior vice president of sales and channel and Jonah Kowall as senior vice president of product and design. The hires come as the company rolls out Spacelift Intelligence, its AI‑focused orchestration...
Propy Secures $100 Million Credit Facility to Build AI‑Powered Blockchain Closing Platform
Propy, the Miami blockchain startup, closed a $100 million credit facility with Metropolitan Partners Group to develop an AI‑enabled, end‑to‑end closing platform on public blockchain. The funding aims to consolidate title and escrow services, speed transactions, and reduce fraud in real‑estate...
DocuSign Teams with Harvey to Embed AI Legal Reasoning in E‑Signature Platform
DocuSign has partnered with Harvey to integrate the AI platform’s legal‑reasoning engine directly into its e‑signature suite. The joint solution lets legal teams pull contracts from DocuSign, have Harvey analyze them against jurisdictional law, and trigger workflow actions without leaving...
ServiceNow Launches Agentic AI Suite, Exposing HR‑IT Priority Clash
ServiceNow introduced its latest agentic AI tools, including Otto and EmployeeWorks, at the Knowledge 2026 conference, revealing a split between IT’s drive for efficiency and HR’s focus on compliance. Executives from CVS Health and analysts warned that misaligned expectations could...
AI Institute Becomes Microsoft Training Services Partner to Scale AI Skills in Africa
The AI Institute has signed a strategic agreement with Microsoft, becoming an official Microsoft Training Services Partner. Launched at GITEX Africa 2024 in Marrakech, the deal will embed Microsoft’s AI curriculum and certifications into the Institute’s programs, targeting the continent’s...
7 Steps to Building Your First AI Assistant
The post walks readers through a seven‑step framework for building a custom AI assistant, illustrated by a solo insurance broker who reduced morning email triage to an automated workflow. It stresses defining a single, well‑documented task before prompting an LLM,...
Datadog CEO's Two‑Word Warning Highlights AI‑Driven Race in Enterprise Monitoring
Datadog chief executive Olivier Pomel warned enterprises they "can't afford to be late" in the AI monitoring race, a message that coincided with a 30% share surge and first‑quarter revenue of $1.006 billion, up 32% year over year. The company also...
Enphase Energy Unveils 1.25 MW Solid‑State Transformer for AI Data‑Center Racks
Enphase Energy announced its IQ Solid‑State Transformer (IQ SST), a 1.25 MW rack‑scale power block built from 342 semiconductor modules. The system promises 98.5% efficiency, 99.999% availability and the potential to eliminate traditional UPS batteries in AI‑focused data centers.
LLMs May Not Overtake Traditional Systems of Record
I don’t fully agree here. Systems of record defensibility has always been around workflows and integrations not just data store. While it’s true that agents can raid them for data - as can any workflow or API call it remains an open...
NHTSA Launches Probe Into Uber‑Partner Avride After 16 Robotaxi Crashes
The U.S. National Highway Traffic Safety Administration opened a formal investigation into Avride, Uber's robotaxi partner, after 16 crashes were reported, one of which caused a minor injury. Regulators say the incidents stem from deficiencies in Avride's autonomous driving system,...
BILL to Cut 30% of Workforce to Accelerate AI Strategy
BILL announced it will reduce its headcount by 30% before the end of the current quarter, making artificial intelligence its top priority. CEO René Lacerte said the move follows rapid adoption of AI agents that have automated over a million...
SAP Blocks External AI on Customer Data, Reinforces Ownership Split
SAP just told customers they can't point external AI agents at SAP data without permission. Throttling, suspension, termination if they do. How could they possibly lay claim to their customers' data? It seems preposterous. Whether through Salesforce's taxation or outright...
Apple Could Unlock iPhone Value by Acquiring Wispr Flow
Apple built this whole buggy Apple Intelligence thing and all they needed to do to make the iPhone infinitely more valuable was acquire Wispr Flow
Nvidia's China Market Share Falls to Zero as Homegrown AI Chipmakers Accelerate
Nvidia CEO Jensen Huang announced that U.S. export restrictions have eliminated the company's direct AI chip sales in China, dropping its market share to zero. Chinese firms such as Huawei, Cambricon, Moore Threads and MetaX are rapidly scaling their own...
Claude Code Stalls without Feedback, Needs Better Status Signals
Sometimes when I message Claude Code it just hangs for 3 minutes and I have no idea whether it's still working or not. Wish it communicated more. https://t.co/Fzd7LhLsYK
AI Amplifies Agency: Winners Gain, Losers Lose
It was always the case that agency was self-compounding, but AI is magnifying the effect. Low-agency AI users further lose agency, high-agency AI users further gain agency.
Teradata Unveils Autonomous Knowledge Platform to Scale Agentic AI
Teradata announced the Autonomous Knowledge Platform, a unified suite that combines AI, analytics and data management to move AI agents from pilot to production. The platform is available now via Teradata Cloud and as an on‑premises Teradata Factory offering, targeting...
Calling Out AI Slop Without Burning Bridges
How do you tell someone that everything they are building and publishing is AI slop?
Super‑intelligence Implies Super‑ethical AI; I'm More Confident There
I love this. If we can build super-intelligence, we can build super-ethical and super-moral AIs as well. Indeed, I'm more confident of the latter than the former.
Shopify Posts 34% Revenue Jump to $3.17B as AI Fuels Growth, Shares Slump 33% YTD
Shopify reported first‑quarter revenue of $3.17 billion, up 34% year over year, and a 35% rise in gross merchandise volume, thanks to AI‑driven features like Sidekick. The earnings beat came as the company's shares have tumbled roughly 33% in 2026, sparking...
AI Promises Stronger Digital Security Despite Short‑term Challenges
It continues to look to me like AI will make the digital world more secure on net and in the long term. The short term may be tricky, but it seems to be going okay.
AI Agents Expose Limits of Traditional Database Design
Great article on how AI agents challenge some fundamental assumptions about databases. Databases Were Not Designed For This https://t.co/CP4szQTgPR
European Startup Humanoid Unveils AI ‘Brain’ Cutting Robot Learning Time to Days
European robotics startup Humanoid announced an artificial‑intelligence system that reduces the time required for humanoid robots to learn new physical tasks from months to just a few days. The breakthrough, described as a unified ‘brain’, could accelerate adoption of robots...
Generative AI Adoption Outpaces Earnings Impact: Operating-Model Gap
Eight in ten companies have deployed generative AI. Eight in ten report no material impact on earnings. That's an operating-model gap. Not an adoption gap. Three days at SAS Innovate. Three moves. In order. https://t.co/uDgZ3K8Z9u https://t.co/EHsM1F5ms1
GitHub Spec Kit Merges Vibe Coding with AI Discipline
GitHub's Spec Kit is a signal: vibe coding and SDD may be converging into a single AI coding environment. The question is whether your team is building the discipline to use it well. #VibeCoding #DevOps #AI https://t.co/gIlTmpGkys
Planning and Evaluation Methods in LLM-Based Autonomous Workflow Systems: A Comprehensive Review
The paper delivers a systematic review of planning and evaluation techniques for large‑language‑model (LLM) driven autonomous workflow systems. It introduces a five‑category planning taxonomy and a five‑axis evaluation framework that measures utility, efficiency, quality, robustness, and safety. The analysis uncovers...
Free 1‑Trillion‑Parameter Model Undermines Paid AI Agents
A Chinese lab just dropped a 1 TRILLION parameter thinking model. For free. It's called Ring-2.6-1T from InclusionAI and it just made every $200/month "agent" subscription look like a scam. Here's why this changes everything ↓ https://t.co/j4a0T8Ddlh
TSMC’s April Revenue Growth Slows to 17.5%, Highlighting AI‑Chip Supply Constraints
Taiwan Semiconductor Manufacturing Co. posted a 17.5% rise in April revenue to $13.1 billion, the slowest monthly expansion since October. The slowdown underscores tightening supply for AI‑focused chips that power enterprise cloud and data‑center workloads.
Morse Code Hack Shows LLM Protections Are Futile
An agent got robbed via Morse code this week. @kaiynne on why this kind of attack is structurally different and why 'protections' on LLMs mean almost nothing once someone wants in. @tayvano_ @LucaNetz @kelsiemvn: https://t.co/WgLDsMJWuj
ChatGPT’s Free Version Is 26 Times More Likely to Respond Inappropriately to Psychotic Delusions
A JAMA Psychiatry study tested three versions of OpenAI's ChatGPT with 79 prompts that mimic psychotic symptoms. The free tier was 26 times more likely to produce inappropriate replies than control prompts, while the paid GPT‑5 version was still eight...
Intent-Based Chaos Testing Is Designed for when AI Behaves Confidently — and Wrongly
Enterprises deploying autonomous AI agents risk catastrophic missteps when the system encounters unfamiliar conditions. A production observability agent mistakenly rolled back a service after flagging a benign batch job, causing a four‑hour outage despite the model behaving as trained. The...