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
Five Ideas for Leading with AI From Gloria Steinem's Living Room
A recent "talking circle" held in Gloria Steinem’s living room gathered top women leaders to distill five practical ideas for leading with AI. The discussion emphasized mapping employee AI readiness, creating safe peer‑learning environments, confronting workflow discomfort, protecting time for AI skill‑building, and anchoring AI to purpose rather than mere efficiency. It highlighted the critical role of change management and women’s perspectives in shaping humane AI adoption. The insights arrive as companies like Meta monitor employee work to train AI agents and firms grapple with benefit cuts and evolving talent strategies.
A Stanford Study Found 38% of Americans Are Excited About AI versus 84% in China, with U.S. Trust in AI...
A Stanford Human‑Centered AI study shows only 38% of Americans are excited about AI, compared with 84% in China and roughly 80% across Southeast Asia. Trust in U.S. government regulation of AI is just 31%, the lowest among surveyed nations....
AWS AI Push Forces CROs to Rethink Scaling and Go‑to‑Market Strategies
Amazon Web Services announced a $200 billion AI‑focused capital spend and a 24% revenue surge, prompting chief revenue officers to overhaul sales organization, scaling models and go‑to‑market playbooks. Executives cite the need for new delivery frameworks as AI services become core...
DescrybeLM Beats ChatGPT, Claude, Gemini on Bar Exam Benchmark
Descrybe unveiled its purpose-built legal AI, DescrybeLM, which correctly answered every question on a 200‑item multistate bar exam benchmark, surpassing leading general‑purpose models. The result highlights the advantage of domain‑specific training for legal reasoning tasks.
Meta AI Veteran Launches New Work Foundation Nonprofit to Arm Gen Z Against AI‑driven Job Loss
Clara Shih, former head of business AI at Meta, has founded the New Work Foundation, a nonprofit that offers AI‑powered tools to help Gen Z navigate a job market reshaped by automation. The initiative debuts platforms like Field Report and...
SEBI Chief Calls for Vision‑Led Tech Framework as Indian Markets Face Disruption
At SEBI’s Foundation Day, Chairman Tuhin Kanta Pandey announced a push for a vision‑led, technology‑centric regulatory framework to safeguard India’s $120 billion annual capital formation market. He highlighted AI‑enabled supervision, e‑office migration and data‑analytics capacity as core pillars.
Cohere to Acquire Aleph Alpha in $20 B Deal, Backed by $600 M Schwarz Group Funding
Cohere announced a deal to acquire Germany’s Aleph Alpha, forming a transatlantic sovereign AI entity valued at roughly $20 billion. The transaction is backed by €500 million (≈$600 million) in structured financing from Schwarz Group, which will also lead Cohere’s upcoming Series E round.
ChatPlayground AI Offers Lifetime Access to Unified 25‑Model Testing Platform for $74.97
ChatPlayground AI introduced a lifetime Unlimited Plan priced at $74.97 (regular $619), consolidating over 25 leading large‑language models into a single workspace. The deal, available through May 10, promises unlimited queries, prompt‑engineering tools, and a Chrome extension, targeting developers and product...
JPMorgan Cuts India to Neutral, Boosts Asian Tech Amid AI Surge
JPMorgan has downgraded Indian equities to a Neutral rating and increased its allocation to Asian technology, especially Taiwan, as AI‑driven market dynamics accelerate. The bank cites rising inflation, earnings pressure and India's limited exposure to next‑gen tech as key reasons...
Rick Vanzura Joins GrowthFactor to Lead Revenue as AI Retail Expansion Gains $5M Funding
Rick Vanzura, the ex‑CEO of Wahlburgers, has been hired as the revenue leader of Cambridge‑based GrowthFactor, an AI platform that advises retailers on store locations. The hire coincides with a fresh $5 million venture round from Teamworthy Ventures and Red Forest...
OpenAI Says Old Prompts Are Holding GPT-5.5 Back and Developers Need a Fresh Baseline
OpenAI’s new prompting guide for GPT‑5.5 urges developers to discard legacy prompt stacks and start with minimal, outcome‑focused instructions. The guide re‑elevates role definitions at the top of the prompt hierarchy and recommends a structured schema covering personality, goal, constraints,...
CFC Launches Lane Assist AI Pilot, Automating Specialty Underwriting in Seconds
CFC announced the launch of Lane Assist, a world‑first AI pilot that automates low‑complexity specialty underwriting from email receipt to quote recommendation in seconds. The move challenges traditional underwriting workflows and could force carriers to accelerate automation or risk losing...
Geely Unveils China’s First Purpose‑Built Robotaxi, EVA Cab, at Auto China 2026
Geely Auto Group, together with AFARI Technology and CaoCao Mobility, revealed the EVA Cab robotaxi prototype at the Beijing Auto Show on April 24, 2026. The vehicle features a quantum‑grade AI electronic architecture, a 2,160‑line digital LiDAR and over 3,000 TOPS of in‑vehicle...
Alphabet to Report Q1 2026 Earnings as CFO Ashkenazi Previews $175B‑$185B AI‑Driven Capex
Alphabet is set to release its first‑quarter 2026 earnings on April 29, with CFO Anat Ashkenazi expected to outline a $175‑$185 billion capital‑expenditure program aimed at scaling AI infrastructure. Investors will watch cloud revenue growth, depreciation pressure and the CFO’s allocation...
Snowflake Unveils AI Platform Upgrade to Serve as Enterprise Control Plane for Agents
Snowflake launched an AI platform upgrade that integrates Snowflake Intelligence and Cortex Code, positioning the data cloud as a control plane for AI agents. The upgrade adds direct hooks to everyday tools such as Gmail, Salesforce and Slack, and supports...
Youtility Secures $4.2 M Seed Funding to Power Demographic‑Free Behavioral AI Targeting
Youtility Decision Science announced a $4.2 million seed round led by Silver Falcon Capital, Fyllo and National Media. The funding will accelerate its behavioral AI platform that replaces traditional demographic segments with real‑time decision‑state modeling, a shift that could upend how...
Anthropic's Mythos AI Finds 2,000+ Zero‑Day Bugs in Seven Weeks, Sparking Safeguard Debate
Anthropic’s Mythos AI model identified more than 2,000 previously unknown software vulnerabilities in just seven weeks, prompting the company to limit access to a handful of trusted partners and spurring regulators worldwide to convene panels on AI‑driven cyber risk. The...
How to Audit What ChatGPT Knows About You - and Reclaim Your Data Privacy
OpenAI provides multiple consumer‑focused controls to limit the personal data ChatGPT retains. Users can opt out of model training, delete chat histories, employ temporary chats, manage or disable memories, and even delete their entire account via the privacy portal. Deleted...
Foundation Tests Humanoid Robot Soldiers in Ukraine, Securing $24 Million Pentagon Contract
Foundation, a robotics startup, conducted a closed‑pilot demonstration of its Phantom humanoid robots in Ukraine, showing they can retrieve supplies under fire. The test follows a $24 million contract with the U.S. Pentagon and has ignited debate over the future of...
AI's Real Bottleneck: Energy, Not Model Size
AI is not just a compute story. It’s an energy story. A few facts: - Data center emissions may be ~6–8× higher than reported - AI tasks can use 10×–70× more energy than traditional computing Demand is accelerating fast. But...
AI’s Power: Databases as Memory
The article argues that pairing large language models with databases transforms AI from a generic knowledge engine into a specialized, reliable enterprise tool. Retrieval‑augmented generation (RAG) and vector search anchor AI answers in verified data, eliminating hallucinations. Databases provide long‑term,...
75% Google Code Now AI-Generated, Says Sundar Pichai: How Company Is Putting AI to Work
Google CEO Sundar Pichai announced that 75% of all new code written at the company is now AI‑generated, up from 50% last fall. The shift to "agentic" workflows lets engineers coordinate autonomous AI agents, accelerating projects such as a complex...
Tolaria: Open‑Source AI‑Ready Markdown Wiki for Mac/Linux
🚨 @Karpathy's LLM wiki concept just became a real Mac app. It is called Tolaria, a free desktop app for Mac and Linux. The goal is simple: give humans and AI agents a shared, native environment to build knowledge. It writes like a...
Keep Engineers, Add AI: Cost‑saving Myth Debunked
The CFO cut 12 engineers to save $1.8M. The AI token bill just hit $2.3M. And now they need engineers to figure out why the AI isn’t working. THE SAVINGS MYTH is the most expensive strategy in tech right now. You didn’t reduce headcount....
Thailand Pilots AI‑Enabled Custom Fertiliser Programme To Reduce Costs For Farmers
Thailand’s Ministry of Higher Education, Science, Research and Innovation and the Ministry of Agriculture have launched a six‑month pilot that uses AI to create farm‑specific fertilizer blends. The system, built by the Thailand Institute of Scientific and Technological Research, combines...
AI Fails From Weak Foundations, Not Lofty Visions
Everyone wants the AI penthouse. Almost nobody wants to pay for the basement. What I keep seeing is the same pattern: companies want AI outcomes without investing in AI foundations. The exciting layer gets funded first: → GenAI pilots → strategy decks → dashboards → executive demos The...
Opus 4.7 Outperforms 4.6 in 1M
Claude Code benchmarks of 200 headless Claude Code sessions comparing Opus 4.6 and Opus 4.7 1M-context models across effort levels and prompt steering variants - concise, step by step, ultrathink
AI Agents Aren't Replacing Software Engineering but Expanding It Far Beyond Code, Researchers Argue
Researchers from Chalmers University and Volvo argue AI agents are not replacing developers but expanding software engineering into "semi‑executable artifacts" such as prompts, workflows, policies, and governance rules. They introduce the "Semi‑Executable Stack," a six‑ring model that moves from core...
Karnataka Drafting Responsible AI Framework To Embrace ‘I-Governance’: Priyank Kharge
Karnataka has formed a responsible‑AI committee to draft a policy framework that will guide the state’s transition from traditional e‑governance to “i‑governance,” an AI‑driven public‑service model. The pilot phase includes semi‑autonomous chatbots for grievance handling, image‑verification tools for crop loss...
AI's Erosion of Privacy
A recent Slate podcast highlighted how individuals are uploading highly sensitive medical data to AI chatbots, treating them as informal therapists despite the lack of privacy safeguards. The episode featured a woman sharing her lifelong diagnoses with ChatGPT and a...
UK Departments at Odds over Energy Demands of AI Datacentres
The UK’s Department of Science, Innovation and Technology (DSIT) forecasts that AI‑focused datacentres will require at least 6 GW of electricity by 2030, while the Department of Energy Security and Net Zero (DESNZ) estimates the entire commercial services sector will grow...
Robots Handle Routine, Freeing Humans for Creative Thinking
Robots are getting better at more than just repeating tasks. Seeing one sort packages, adjust on the fly, and keep going even after something breaks is pretty impressive. This isn’t really about replacing people. It’s about taking over the repetitive stuff so humans...
Self‑Driving Cars Fail Where Human Negotiation Matters
A few self-driving taxis in San Francisco just demonstrated the real problem with autonomy. They were too rational. For a brief moment, several robotaxis aligned at an intersection and created a perfectly polite deadlock. No aggression. No improvisation. No human-style “you go, I’ll go.” Just algorithms...
Deepvue Overview: Quick Start, AI Tools, Time Saver
If you are interested in @Deepvue this video gives a great overview of the platform and how it can save you time ✅ Quick Start Guide ✅ Top Screen Presets ✅ Deepvue AI Features ✅ Theme Tracker & Dashboard https://t.co/WwhuZzUsGT
Conversation Routines Introduce AI to Young Learners
📚 Want to introduce AI concepts to younger students in a developmentally appropriate way? 💬 Dr. Nneka McGee shares conversation based routines that help students build foundational understanding. 🎧 You’ll leave with ideas you can try without logging into anything. https://t.co/EJdIQUi09m
AI Augments Abstractors, Saving 30 Minutes per Case
A healthcare data abstractor feared AI would replace her. Instead: AI caught data she was missing. 30 minutes saved per case. This week, Stakeholder Engagement is up 100%. Collaboration up 29%. The skills that matter are changing. Not disappearing. https://t.co/YhNGf2OEhE
Hit 1 Billion Tokens Locally, Saving £750
Personal landmark as I crossed a billion tokens used on my local LLM since setting it up just over a month ago. If I'd run this through API calls on big tech's models, it'd have cost me around £750 or...
AI Turns Decades of Mastery Into Weeks
The cost of trying something has dramatically decreased. So experience that took decades to accumulate can now be speed run with the help of AI. Change is cheaper Failure is cheaper Learning is faster So young people can get world class...
Startup Sells AI Clones of Human Experts
This Startup Wants You to Pay Up to Talk With #AI Versions of Human Experts by @StevenLevy @WIRED Learn more: https://t.co/xEyxpv2eT9 #ArtificialIntelligence #MachineLearning #ML https://t.co/pZiQVE7XZy
If AI Writes Better, You're Not a Copywriter
I'm hearing a lot of copywriters saying AI writes better than them. Boy, if AI is writing better copy than you, then you really aren't a good copywriter.
AI Experts Optimistic, Public Fears Job Losses
There's a growing gap between AI experts and public opinions: 73% of experts but only 23% of the public are positive about the long-term effect on jobs, "with nearly 2/3s of Americans thinking that AI would lead to fewer...
Managing Hidden Risks of Enterprise Generative AI
Governing the hidden risks of generative #AI in the enterprise by Ian Jeffs @techradar Learn more: https://t.co/E4WNyH6G8t #MachineLearning #ArtificialIntelligence #ML https://t.co/UQzYKT0KW7
FSD 14.3.2 Auto‑adjusts to Rain, Keeps Driving Smooth
FSD 14.3.2 performed very nicely in the rain today Took me down to Standard when the rain got heavy, and then bumped me up back up to hurry shortly after as it started to clear up a little bit love...
AI Reads 40,000 Community Posts Daily for You
Why are you trying to read 40,000 posts from the AI community here a day? My AI does that for you: https://t.co/kiuZ7QXLzb
GBrain v0.22 Launches with Search Fixes and Separate Eval Repo
Releaseing GBrain v0.22 - lots of fixes to search and retireval and a new eval system (gbrain-evals is now a separate repo to prevent checkout bloat) https://t.co/qVAQjQeReV
Control Your Browser Directly with Claude Code
Did you ever want to control your browser side-by-side with Claude Code? Now, with /open-gstack-browser skill and GStack Browser, you can https://t.co/dT0dUd3MhF
Self‑Programming Materials Will Make Infrastructure Self‑Repairing
one of the areas i’m watching for @maniacvc: if intelligence embeds into everything, matter becomes self-programmable. roads fix themselves. bridges know when they’re tired. the built world becomes regenerative. it’ll be the end of the era where things break and stay broken.
AI Mastery Requires Guardrails to Prevent Burnout
AI fluency is super important but none of it matters if you can’t put in guardrails to help people prevent burnout. It’s the only thing I’m concerned about with people who ‘get AI’. The machine is overclocked but must stay cool to...
Every Decade Redefines Business: From Internet to Cybersecurity
2000s: every company is an internet company 2010s: every company is a software company 2020s: every company is an AI company 2025+: every company is a cybersecurity company
US State Dept Issues Global Alert on Chinese AI Theft
Exclusive: US State Dept orders global warning about alleged AI thefts by DeepSeek, other Chinese firms https://t.co/pddPI6zgPF