Today's AI Pulse
Anthropic shuts down Claude Fable 5 and Mythos 5 after US export‑control warning
Anthropic disabled access to its Claude Fable 5 and Claude Mythos 5 models after the Trump administration, via the Commerce Department, warned that foreign nationals could exploit the systems. The move followed alerts from Amazon and other firms about a potential jailbreak that could aid cyber‑security attacks. Anthropic declined the administration’s request for a voluntary shutdown but complied with the order.
Also developing:
By the numbers: AMD acquires AI memory‑optimization pioneer Mext
The Next Phase of AI Spending Is Already Underway
The AI spending cycle is moving from a fast‑paced experimentation phase to a disciplined, production‑focused era. Gartner forecasts total AI spend of $2.5 trillion in 2026, with $401 billion earmarked for infrastructure. Enterprises now confront surprise operating bills as inference, data movement, and token usage drive costs far beyond initial pilots. Investors are being urged to shift attention from chip makers to the firms that solve MLOps, data‑modernization and AI‑cost‑optimization challenges.
Genius Sports Integrates Multi-Product Portfolio Across LigaMX Data Operations
Genius Sports has sealed a multi‑product partnership with Mexico’s top‑flight Liga MX, rolling out its GeniusIQ data and AI platform across all league stadiums. The suite powers real‑time advertising, augmented broadcasts, semi‑automated offside (SAOT) technology, and a full analytics suite for...
AI Arms Race Shifts to Securing Data Centers
The Next AI Arms Race Is About Fortifying Data Centers https://t.co/FgUFDOP3m7 via @YouTube #AIarmsrace #AIweapon #datacenter #storage #artificialintelligence #AI @lexfridman @KirkDBorne @Ronald_vanLoon @erikbryn @antgrasso @sallyeaves @Nicochan33 @HaroldSinnott @mvollmer1 @marcusborba @CatherineAdenle @vinod1975 @YuHelenYu @ingliguori @fogoros @ylecun @TerenceLeungSF @SourabhSKatoch @wil_bielert @HsrYueli @faustospain @Whats_AI...
Inworld AI Launches Realtime TTS-2: A Closed-Loop Voice Model That Adapts to How You Actually Talk
Inworld AI unveiled Realtime TTS‑2, a research‑preview voice model that ingests the actual audio of a conversation rather than just text transcripts. The closed‑loop architecture lets the system detect tone, pacing and emotion from prior turns, enabling more natural, context‑aware...
Dreame Unveils All‑in‑One Smart Yard Hub and Three AI‑Powered Garden Robots in San Francisco
Dreame Technology rolled out an All‑in‑One smart‑yard centre, the A3 AWD PRO robotic mower and the APEX multifunction robot at its Dreame NEXT: Autonomous Outdoors showcase in San Francisco on April 29. The launch signals the Chinese firm’s bid to...
Ulta Beauty’s AI‑Driven Rewards Program Generates 95% of Sales
Ulta Beauty’s chief marketing officer Kelly Mahoney says the company’s AI‑enhanced rewards program now fuels 95% of its sales, drawing on data from 46.7 million members. The move highlights how AI‑driven personalization can turn loyalty programs into the core revenue engine...
Cisco to Acquire Israeli AI‑Security Startup Astrix for Up to $400 Million
Cisco Systems will purchase Israeli cybersecurity firm Astrix Security in a transaction valued between $300 million and $400 million. The deal adds Astrix’s platform for discovering, governing and securing AI agents and non‑human identities to Cisco’s security suite, addressing a growing blind...
AI Lets Chemists Design Molecules by Simply Describing Them
Researchers at EPFL introduced Synthegy, an AI framework that pairs large language models with traditional retrosynthesis software. Chemists can input plain‑language instructions, and the system scores and explains generated synthetic routes, aligning them with strategic goals. In a double‑blind test,...
Kalshi Rolls Out Facial-Recognition Safeguards to Bar Under‑age Traders
Kalshi announced a suite of new identity‑verification tools—including facial recognition, selfie checks and two‑factor authentication—to keep users under 18 out of its prediction‑market platform. The measures arrive as lawmakers and sports leagues push for stricter age limits on derivative‑style contracts.
Bregal Milestone Takes Majority Stake in Meteoviva, Boosting AI Energy Management
Bregal Milestone announced a majority growth investment in meteoviva, the AI‑powered building energy‑management platform, to fund rapid European expansion. The deal gives meteoviva access to capital and operational support while the private‑equity firm taps a market where AI and decarbonisation...
Converge Bio’s AI Platform Doubles Cetuximab Affinity in Eight Hours
Converge Bio announced that its generative‑AI platform ConvergeAB produced a cetuximab antibody with more than double the binding strength to EGFR in an eight‑hour, zero‑shot experiment. The result, filed as a provisional patent, highlights AI’s capacity to accelerate and refine...
Sage Unveils "Glass Box" AI Suite for Finance and HR, Adding Explainable Tools to Mid‑Market
Sage announced a new "glass box" artificial‑intelligence suite at its Sage Future 2026 conference, coupling explainable AI with finance and human‑resources software. The rollout includes the Sage HCM platform for mid‑market firms, the acquisition of Doyen AI, and deeper collaborations...
DDR6 Server Memory Enters Early Development to Power Next‑Gen AI Workloads
Memory manufacturers and key supply‑chain partners have started early hardware validation of DDR6 server‑grade DRAM. The move, announced ahead of any formal JEDEC standard, signals the industry’s push to satisfy exploding AI compute requirements.
Incredibuild Unveils Islo Sandbox to Secure AI Coding Agents at Scale
Incredibuild announced the launch of Islo, a sandbox that isolates AI coding agents in dedicated cloud environments with granular security policies. The solution lets enterprises run agents 24/7 while protecting codebases, secrets, and infrastructure, aiming to turn experimental AI workflows...
OpenAI Fast‑tracks AI‑first Smartphone for 2027 Launch, Taps MediaTek
OpenAI is accelerating development of its first AI‑first smartphone, planning mass production in the first half of 2027. The device will run a customized MediaTek Dimensity 9600 processor and could ship up to 30 million units across 2027‑28, signaling a bold...
Instantly See Which AIs Can Access Your Site
Instant insights. in seconds you'll know which AIs can access your site ...and which cannot.
AI Overcapacity Fuels Self-Preferencing, Undermines Real Sense‑Making
Recruiting Brainfood - Issue 499 https://t.co/Gjptm9PGe7 AI Self-preferencing, Overcapacity of AI, one insane essay on sense making models for organisations when nothing out there is real and playing games being a great spend of your time... https://t.co/mfMP1PeGNy
GenAI Already a Major Component of Care Delivery
A CSA Group report authored by Will Falk argues that Canada must accelerate supervised adoption of generative AI in healthcare. The paper cites current clinician use for documentation and patient‑facing tools for explanation and navigation. It warns that delaying implementation...
Enterprise AI Needs Real Context, Not Marketing Hype
Some areas of coding AI are super human - Ethan Mollick Every term in AI is made up, taken over by marketing, slowly becomes reality. What matters now is giving the AI the context of the enterprise. #AtlassianTeam26 #AI #CIO https://t.co/SjODPJIolU
AI's
Sometimes AI is super human, other times it gets stuck - long form fiction writing for example. Its performance is jagged. - Ethan Mollick AIs as a native NY driver - “very unpolite” - Magnus Ostenburg #AtlassianTeam26 https://t.co/tZlp7r5j7B
Tesla Hits 10 Billion Miles, Meeting Musk’s ‘Safe Unsupervised’ Driving Threshold
Tesla announced its fleet has driven more than 10 billion miles, satisfying Elon Musk’s benchmark for safe unsupervised self‑driving. The milestone fuels speculation about a shift to fully autonomous FSD, while legal and regulatory questions linger.
Experience Trumps 'AI‑Native' Labels; Keep Exploring
AI native doesn’t mean very much. Experienced people are better at AI than juniors. AI is a moving target. Nobody knows what’s going on with AI so exploration is really important. Use it for actual work but alsl use your...
KNEX Technology Launches AI Agent Pack for PeopleSoft to Automate HR Tasks
KNEX Technology introduced its AI Agent Pack for PeopleSoft, embedding three domain‑specific AI agents that automate payslip queries, benefits navigation and time‑off planning. The launch aims to cut administrative load and improve employee experience without requiring system overhauls.
Insurity Challenges AI Hype in Insurance Core Systems, Calls on Insurance Carriers to Demand Real Cost and Timeline Reductions From...
Insurity, a cloud‑software leader for P&C carriers, is pushing back against the wave of “agentic AI” hype from legacy core‑system vendors. It argues that most AI add‑ons merely sprinkle intelligence into narrow tasks while still demanding costly, multi‑year professional‑services engagements....
Banza Secures $1 Million Pre‑Seed to Build Personal AI Twins
Banza, an Indian AI startup, raised $1 million in a pre‑seed round led by Campus Fund, with participation from Avalanche and angel investors. The funding will accelerate development of its personal AI “Twin,” a cross‑platform assistant that learns a user’s preferences...
Databricks Invests AUD $420 Million to Expand Data‑lake and AI Services Across ANZ
Databricks announced a AUD 420 million (about US$280 million) three‑year investment in Australia and New Zealand, adding a 22,000‑sq‑ft Sydney headquarters and scaling its Lakebase, Genie and Agent Bricks products. The plan also includes training 100,000 learners, reflecting more than 85% YoY regional growth.
US VC Funding Value Jumps Over 200% in Q1 2026, Mega‑Deals Lead Surge
U.S. venture‑capital funding value surged more than 200% in the first quarter of 2026, even as the number of deals rose only 5%. The jump was powered by a handful of AI‑focused mega‑rounds, underscoring a shift toward late‑stage capital concentration.
SaaS ERP Market Set to Hit $281.6 B by 2034 as AI and Cloud Drive Growth
The global SaaS‑based ERP market is projected to reach $281.58 billion by 2034, expanding at a 13% CAGR as AI‑enabled, cloud‑native platforms become the norm. The surge is driven by enterprise demand for built‑in automation, the dominance of SMEs, and rapid...
China Deploys Embodied AI Robots for Hazardous Industrial Work
RobotPlusPlus unveiled a 90‑kg climbing robot that can weld, scan and spray on vertical steel tanks, while a new subsea cable‑inspection robot and grain‑leveling machines expand China’s embodied AI push. The moves reflect the country's five‑year plan to make AI‑enabled...
Agents Write Code. Fixing It Is Still On You.
Amplitude’s new MCP platform embeds session replay data directly into AI coding assistants like Claude and Cursor, allowing developers to investigate bugs without leaving their code editor. By describing an issue in plain language, the system triggers specialized skills—debug‑replay, diagnose‑errors,...
F5 AI Guardrails Quickstart: Answering the Hard Questions
A financial services firm is piloting an AI assistant that draws answers from its own underwriting manuals and regulatory filings. To address security gaps, F5 offers an AI Guardrails quickstart that adds an inline inspection layer for both prompts and...
AI, Strategy, and the Future of Work: Oxford Economist Jean-Paul Carvalho
Oxford economist Jean‑Paul Carvalho argues that artificial intelligence is reshaping corporate strategy and the future of work. He notes that AI can lift productivity by as much as 30% when embedded in core processes, but the gains hinge on addressing...
From Efficiency Economy to Judgment Economy: Cases&Faces Conference on the New Value System in Business
The Cases&Faces Conference on May 11, 2026 highlighted a transition from an efficiency‑driven economy to a judgment economy, where AI tools are ubiquitous and competitive advantage stems from human insight. Speakers from Google, Nike, Coca‑Cola and others emphasized asking the...
SiTime’s Latest TCXOs Boost GPU Utilization in AI Data Centers
The request references an article titled “SiTime’s Latest TCXOs Boost GPU Utilization in AI Data Centers,” but the full text of that article was not included in the provided material. Without the article’s content, specific details about the new temperature-compensated...
GPT-5.5 Instant Shows You What It Remembered — Just Not All of It
OpenAI has made GPT‑5.5 Instant the default model for ChatGPT, replacing GPT‑5.3 Instant and promising higher dependability, accuracy, and reasoning. The rollout introduces “memory sources,” a UI feature that surfaces which prior chats or uploaded files the model consulted when generating a...
Anthropic to Invest $200B in Google Cloud and Chips
Anthropic will reportedly spend $200 Billion with Google $GOOGL over the next 5 years on Cloud and Chips - The Information
Free GPT‑5.5 Instant Matches Paid 2025‑level Performance
All benchmarks are flawed, but GPQA has been fairly consistent & highly correlated with other measured benchmars. I think it's a good way to see how far we've come that the free model from OpenAI, GPT 5.5 Instant, is at...
Mistral Medium 3.5 Is Live in Kilo Code
Mistral announced the public preview of its Medium 3.5 model, a 128‑billion‑parameter dense LLM with a 256k token context window, now live on the Kilo platform. The blended architecture unifies instruction‑following, reasoning and code generation, achieving a 77.6% score on SWE‑Bench,...
AI-Enabled Vulnerability Discovery Is Reshaping National Cyber Defence
Anthropic’s Claude Mythos LLM demonstrated strong vulnerability‑discovery ability, flagging 271 flaws in Firefox during preview testing. AI‑enabled tools now automate key steps of zero‑day exploit development, dramatically reducing time and cost compared with traditional manual methods. The UK warns that...
Claude Content OS: Your Full AI Content Team
Claude Content OS launches as an all‑in‑one AI content team that promises to multiply output while preserving brand voice. The product includes a two‑hour hands‑on workshop, a pre‑configured folder of prompts and slash commands, and 21 plug‑and‑play skills covering writing,...
Gemini Readies Agent to Drive Your Mac and ‘Organize Your Files’
Google is adding a Gemini‑powered agent to its macOS app that can directly control the computer, organize files, and push data into Google Workspace. The agent will use screen‑access and accessibility APIs to see the screen, move the mouse, type,...
Enterprises Contain AI Agents to Balance Risk, Reward
Enterprises are rapidly deploying AI agents but face heightened security and compliance risks, prompting a shift toward contained experimentation. Axos Bank reduced its AI tester pool from 300 to a focused five‑to‑seven‑person team to enforce governance and prevent data leakage....
Claude Starter OS: Your First AI Operations Team
Claude Starter OS is a $9, one‑time purchase that bundles 26 AI commands to automate routine business tasks such as email triage, meeting briefs, and document cleanup. The tool plugs directly into seven everyday platforms—Gmail, Calendar, Drive, Docs, Sheets, Notion,...
Choose Camry‑sized Agents over Ferrari Demos for Real Results
Everyone wants a Ferrari agent. Expensive, beautiful, looks insane in the demo. Then you try to run your business on it and realize you built the AI version of a car that scrapes every driveway in LA. The temptation is obvious. You bolt...
Mythos Scores Just Above AI Trend, Not Revolutionary
Another new analysis finds that Mythos, while a fantastic model, is either pretty much on the AI trend or just slightly above it. [Similar to what I found in normalizing scores on Epoch's capabilities index.] https://t.co/KTwEdWM4Q3 https://t.co/33cn9vxSl4
Google Readies ‘AI Ultra Lite’ Plan and Explicit ‘Usage Limits’ for Gemini
Google is preparing a new “AI Ultra Lite” subscription tier that sits between its $20 Pro and $250 Ultra plans, adding a dedicated dashboard for token‑budget monitoring. The move responds to a growing token crunch as AI agents and coding...
Google Gemini Expands to Cars, Sparking Privacy Concerns
Google Gemini is rolling out to millions of cars, but some owners are concerned about data privacy. https://t.co/ffvLwEJQFu
Netflix Routes Millions of Requests to Thousands of Models
Wow. The Netflix ML serving platform serves hundreds of model types/versions, and around 1 million requests per second. How do they route traffic to the right model in such a large-scale service system? Deep dive post: https://t.co/1sU4Jtt5l8 https://t.co/ANPlcxqpSA
Toward a Better Evaluations Ecosystem
Current AI model evaluations suffer from inconsistent methodologies, making headline numbers incomparable across companies. The article cites frequent changes in SWE‑bench setups at Anthropic, OpenAI, and Google, highlighting how tool usage, trial counts, and dataset subsets vary. It proposes a...
True, Google Partner on 'AI for All Thais' Programme
True Corporation Plc has teamed up with Google to launch the “AI for All Thais” programme, backed by Thailand’s Ministry of Higher Education, Science, Research and Innovation. The initiative seeks to democratise AI education through a two‑phase rollout, beginning with...