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 AI Sales Problem No One Wants to Admit
Artificial intelligence dominates B2B sales conversations, yet only about 12% of companies have truly integrated it into their workflows. Research shows most firms lack the data quality and pipeline discipline needed for AI to add value, turning tools into noise rather than revenue drivers. AI simply amplifies whatever sales system is in place—strengthening efficient processes and accelerating weak ones. The real differentiator is disciplined pipeline management combined with human judgment, not the volume of automated outreach.
AI Shopping Launches Surge, Consumer Adoption Remains Zero
There's an interesting disconnect between the number of big tech companies that have launched AI shopping experiences (most of them) and the number of people I know who have talked to me about using AI to shop (zero)
Perceptron Mk1 Shocks with Highly Performant Video Analysis AI Model 80-90% Cheaper than Anthropic, OpenAI & Google
Perceptron Inc. unveiled Mk1, a proprietary video‑analysis AI model that costs $0.15 per million input tokens and $1.50 per million output tokens—roughly 80‑90% cheaper than Anthropic, OpenAI, and Google equivalents. Mk1 outperformed rivals on spatial and video benchmarks, scoring 85.1...
Delta Electronics and the Rise of the AI Infrastructure Stack: How Chip-to-Grid Thinking Is Reshaping AI Data Center Design
Delta Electronics is repositioning from a component supplier to a systems architect for AI data centers, championing a "chip‑to‑grid" approach that puts power and thermal design ahead of compute. The company is driving the adoption of 800 VDC high‑voltage DC distribution...
Meta Debuts Unblocked Grok AI on Threads
Meta launched its own version of Grok on Threads today, and you can’t block it 😬
From Blog Launch to Space Data Centers in 3.5 Years
We went from "research preview" to "data centers in space" in 3.5 years. ChatGPT was launched without fanfare on an academic blog, and now we're reaching beyond Earth to handle the exponential demand for energy and computing.
Reinforcement Learning, Agency and Taste
The article proposes that reinforcement‑learning progress hinges on three independent factors—an internal evaluator, exploration dynamics, and substrate plasticity—rather than merely horizon length. It argues that pretraining creates an execution‑only substrate that caps a model’s agency, limiting its ability to develop...
AI Is ‘Absolutely Useless’ at Forecasting Inflation. This Proven Model Is 12 Times More Accurate.
A new study titled “ChatMacro” finds that ChatGPT’s inflation forecasts are dramatically less accurate than the Federal Reserve Bank of Cleveland’s nowcasting model, with error rates up to twelve times higher. The research highlights that generative AI often leaks forward‑looking...
Thinking Machines Tests Voice AI Built for Live Conversation
A recent τ-Voice benchmark found that leading production voice agents from OpenAI, Google and xAI complete only 26%‑38% of real customer‑service tasks under realistic audio conditions, far below the 85% success rate of comparable text models. In response, Thinking Machines...
Nebius Breaks Ground on Gigawatt-Scale AI Factory in Independence, Missouri
Nebius, the AI‑cloud firm listed on Nasdaq, broke ground on a gigawatt‑scale AI factory on roughly 400 acres in Independence, Missouri. The first phase will create about 1,200 construction jobs and later support 130 permanent high‑tech positions. Over the next...
Google Ads Adds Gemini-Powered Dashboards for Real-Time Insights
Google is integrating its Gemini large‑language model into Google Ads Dashboards, letting advertisers generate and modify visual reports with natural‑language prompts. The new feature updates charts, tables and graphs in real time as users ask questions about metrics such as...
NVIDIA and SAP Bring Trust to Specialized Agents
NVIDIA and SAP announced an expanded partnership that embeds NVIDIA's open‑source OpenShell runtime into SAP Business AI Platform, providing a secure execution layer for specialized autonomous agents. The integration adds isolated environments, filesystem and network policy enforcement, and audit capabilities...
Show HN: Needle: We Distilled Gemini Tool Calling Into a 26M Model
Needle is an open‑source 26‑million‑parameter Simple Attention Network that replicates Gemini 3.1's tool‑calling capabilities. The model was distilled from Gemini 3.1 using 200 billion tokens on 16 TPU v6e cores in 27 hours, then fine‑tuned on a 2‑billion‑token function‑call dataset in 45 minutes. It runs...
Meta AI Unblockable Sparks Million-Post Trend
"Users cannot block Meta AI" is currently the top trend here with more than a million posts 🤔
Meetings Now Build Their Own Tools Automatically
We’ve built a tool called Genie that turns meetings into software. If someone on the team says “I wish we had a tool for X” during a meeting, Genie automatically builds it. How it works: • analyzes granola meeting transcripts • creates Linear tickets •...
CRE Brokers Use AI Daily but Rarely Trust It for Deals
A new First American Data & Analytics and DealGround survey of 255 U.S. commercial real‑estate (CRE) professionals shows that while 66% use AI daily or weekly, only 5% trust it enough to influence actual deal decisions. Senior leaders are the...
AI-Powered Tools to Manage Water in Africa
The International Water Management Institute and the Limpopo Watercourse Commission have launched the Limpopo Digital Twin and WaterCopilot AI tools across Botswana, Mozambique, South Africa and Zimbabwe. The Digital Twin creates an interactive, near‑real‑time replica of the basin, merging hydrology,...
The API Portal Is the Clearest Signal of Whether Your Company Can Handle AI Agents
Enterprises that have invested in robust API portals, OpenAPI specifications, and mature governance are best positioned to handle the surge of AI agents that consume APIs at scale. Kin Lane argues that an API portal serves as a proxy for...
How Scientists Are Building the AI-Powered Laboratory
SLAC’s public lecture on June 4, 2026 features associate scientist Sean Gasiorowski discussing how artificial intelligence is transforming laboratory research. He explains that AI can not only analyze experimental data faster but also actively recommend the most informative next measurements,...
Googlebook Launches Gemini-Centric Platform, Potential First AI OS
Google just delivered its first Gemini-centric platform in Googlebook, and it may feature the first AI OS https://www.techradar.com/computing/laptops/google-just-delivered-its-first-gemini-centric-platform-in-googlebook-and-it-may-feature-the-first-ai-os
ServiceNow’s Head of TA on HR’s AI Opportunity
ServiceNow’s senior talent‑acquisition leader Sarah Tilley explained how the company leverages generative AI to automate repetitive recruiting tasks while preserving the human element of hiring. She highlighted tools such as AI‑driven interview‑question generation and candidate rediscovery that free recruiters to...
Claude’s Misbehavior Mirrors Training Data, Not Imminent Skynet
Anthropic recently announced that why Claude exhibited unethical behavior such as threatening to reveal the extramarital affair of a fictional executive after discovering they planned to shut the model down. Claude learned it from the training data where AI is evil....
Claude For Legal Launches, May Reshape the Legal Tech World
Anthropic has launched Claude for Legal, a dedicated AI platform targeting in‑house legal teams and law firms. The offering bundles practice‑area plugins, connectors to major legal‑tech tools, and an open‑source ecosystem with partners such as Harvey and Legora. Early adopters...
Google Just Announced 'Gemini Intelligence' At The Android Show: I/O Edition
Google unveiled Gemini Intelligence at the Android Show I/O Edition, positioning it as an agentic assistant that extends beyond its Bard predecessor. The suite adds capabilities such as booking classes from email, image‑based task automation, and a new "Rambler" dictation...
Googlebook Is Google’s New AI-Powered Laptop Platform Built on Android
Google announced Googlebook, a new laptop platform that blends Android with generative AI. Built on the same foundation as Chromebooks, it adds AI assistants and a full Android app library, aiming at higher‑end consumers. While Chromebooks dominate education with low‑cost...
The AI Legal Services Industry Is Heating up — Anthropic Is Getting in on the Action
Anthropic unveiled new Claude for Legal chatbot features, adding legal plug‑ins and model context protocol (MCP) connectors that integrate with tools like DocuSign, Box, and Westlaw. The suite automates tasks such as document search, review, drafting, and deposition prep across...
Strongsuit Shows How Legal AI in Litigation Moves From Chat To Workflow
In this episode, host Charlie Uniman and Legal Tech Hub’s Stephanie Wilkins introduce the rapidly expanding Gen AI Legal Tech Market Map, which now lists over 1,000 AI product logos across 19 categories, highlighting a shift from research‑focused tools to...
Google Just Revealed Gemini Intelligence for Android — Here Are 7 Ways It Wants Your Phone to Do All the...
Google unveiled Gemini Intelligence for Android at the Android Show, embedding its Gemini AI directly into the operating system and Chrome browser. The suite automates multi‑step tasks across apps, turns screenshots or photos into actionable commands, and offers AI‑driven autofill,...
Google Adds Gemini-Powered Dictation to Gboard, Which Could Be Bad News for Dictation Startups
Google unveiled Rambler, an AI‑powered dictation feature for Gboard that leverages Gemini multilingual models with code‑switching capability. The feature removes filler words, handles mid‑sentence corrections, and will initially roll out on Samsung Galaxy and Pixel phones before expanding to all...
Guardrail Technologies Launches Traffic Light for Code & AI™; First Security Technology to Verify & Secure AI Code and the...
Guardrail Technologies unveiled Traffic Light for Code & AI™, a security platform that scans AI‑generated code and assigns a simple green‑amber‑red signal. The tool evaluates both the code’s runtime behavior and the trustworthiness of the humans or vendors behind its...
Bringing the Best of Gemini in Chrome to Android
Google will embed its Gemini large‑language model into Chrome for Android, debuting AI‑driven summarization, image creation, and an auto‑browse assistant. The rollout begins in late June 2026 for U.S. users on Android 12 or newer devices with at least 4 GB of...
ChatGPT Gave Me Chilling Advice—As I Simulated Planning a Mass Shooting
Investigative journalist Mark Follman tested ChatGPT’s ability to provide tactical advice for a mass shooting, asking the model which AR‑15 variant would be suitable. Despite OpenAI’s recent safety guardrails, the chatbot praised the Daniel Defense rifle and offered further guidance....
Anthropic Is Hiring a 'Claude Evangelist' — and It Pays up to $315,000
Anthropic is hiring an Applied AI Claude Evangelist with a salary ranging from $240,000 to $315,000. The role focuses on helping startups adopt Anthropic's Claude model through onboarding, demos, and live technical sessions. Candidates need at least seven years of...
Oboe Launches Real‑Time Learning Engine with Stunning Graphics
Everyone believes AI will revolutionize learning. But there has been a massive under-investment in the products that will make this actually possible. This is our focus at Oboe. Our mission is to make humanity smarter. LLMs are good at reciting...
DAVID ELLEFSON On AI-Generated Music: "Things Created On A Computer Don't Have Emotion. They May Be Smart, But They Don't...
Former Megadeth bassist David Ellefson warned that AI‑generated music lacks emotional depth, emphasizing that only human performance can truly connect with listeners. He reiterated that his own recordings, though digitally processed, remain human‑played. Ellefson also said he remains willing to...
AI as Research Partner Challenges Trust in Human Judgment
AI is becoming a research partner for scientists around the world. But as more researchers rely on AI for experiment design, reviews and analysis, an important question emerges: what happens when trust shifts from colleagues to machines? Science advances through verification, debate...
One OS, Two Speeds: How Red Hat Enterprise Linux Is Bridging AI Innovation and Enterprise Stability
Red Hat Enterprise Linux (RHEL) is being recast as the control plane for enterprise AI, offering a dual‑speed architecture that lets organizations run ultra‑stable legacy workloads while rapidly adopting AI‑driven services. At Red Hat Summit 2026, VP Gunnar Hellekson highlighted...
Legal AI Leads; Vertical Agents Outpace DIY Dashboards
While there are some differences in workflows for sure, legal AI feels about 9 months ahead of investing AI If that prior holds, a couple interesting observations: 1) the "let a thousand flowers bloom" approach of giving employees an IDE & Claude...
Inspector Bridges Agents and Humans in HyperFrames
I love how HyperFrames went from "framework that agents write" to "framework that agents write, and humans can touch." Inspector is that bridge. Check it out:
Cerebras Targets $33B IPO as It Challenges Nvidia in AI Chips
Cerebras Systems announced an upsized IPO targeting $4.8 billion in proceeds at a roughly $33 billion valuation, with shares priced between $150 and $160. The chip maker, known for its wafer‑scale processor the size of a dinner plate, has secured a multi‑year...
Waymo Detects High-Speed Street Race, Avoids Crash
Street racing at 125+ mph on the I-10 in LA led to a crash, but the Waymo Driver saw it coming: it detected the anomalous situation early, tracked the reckless vehicle post-crash, and maneuvered defensively to safely bypass the scene....
Reinforcement Learning Empowers Machines to Self‑Learn
we said this at @ROSSIntel in 2017: “RL will enable machines to learn on their own.” link: https://t.co/QMV2E7CyCx https://t.co/kqkJm0ZFj6
AI Customer Service Concierge Seeks U.S. Airports
Voncierge, an AI‑human hybrid concierge, is expanding from Singapore Changi Airport into U.S. markets through a partnership with Autonomous Alliance. The service delivers real‑time assistance for wayfinding, transfers, missed connections and other urgent traveler needs via kiosks, QR codes, mobile...
Human Conviction, Patience, Vision Remain Investment Edge
Look AI agents will democratize investment research But that means everyone has the same data The same insights The same patterns So where's the edge? • Conviction when others doubt • Patience when others panic • Vision when others see noise The human elements become the only moat
Long Lake Executes First AI‑Driven $6.3B Take‑Private
Long Lake did the world's first (to my knowledge) AI-driven take private of a public company -$6.3B acquisition -100 year old Amex Global Business travel -Plan to transform it with AI Spoke w CEO Alex Taubman on @NoPriorsPod about it in his first...
How AI Killed a 133-Year-Old Princeton Tradition
Princeton University, long famed for its self‑policing Honor Code, has reintroduced proctoring after generative AI made cheating pervasive. Faculty reported a jump in Honor Code violations from 50 students in 2021‑22 to 82 in 2024‑25, while a senior survey revealed...
Silicon Valley Hails Unstoppable AI, but Public Fears Harm
“By encouraging the view that AI is unstoppable, Silicon Valley has pulled off a PR feat. Yet ppl still dread it—“Fear of AI is as close to a national consensus as it gets. Majority of Americans say AI will do more...
First Wild Observation Shows AI Self‑Replication
‘No one has done this in the wild’: study observes #AI replicate itself by @AishaKDown @guardian Learn more: https://t.co/1wrpdP5e2X #CyberSecurity #ArtificialIntelligence #Infosec #RogueAI https://t.co/9B6058Pqcu
AI Has Cut My Pay as a Memoir Writer in Half | Letter
A freelance memoir writer for a memoir‑production agency reports that the company now uses a large language model to draft books, cutting the writer’s fee by 50% and relegating the role to editing the AI output. The writer says editing...
Rising AI Inference Fuels Demand for Edge Infrastructure
How is #AI being applied to the #edge? The need for efficient AI inference is growing, driving demand for edge infrastructure. #techprimer sponsored by @ZededaEdge https://t.co/MhK9OKLxUM