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
Mars Rover Drives with the Help of Anthropic AI
Engineers at NASA's Jet Propulsion Laboratory used Anthropic's Claude AI to map a 450‑meter route for the Perseverance rover on Mars. By feeding Claude years of mission data, the system generated Rover Markup Language commands that covered 10‑meter segments and modeled over 500,000 variables. The AI‑assisted process cut route‑planning time roughly in half, with engineers only making minor supervisory tweaks. This demonstration suggests AI could streamline navigation for future deep‑space missions where communication delays are prohibitive.
Nansen Launches Web Version of AI Agent to Improve Onchain Analysis and Trading
Blockchain analytics firm Nansen has launched its AI Agent on a web platform, extending the mobile‑first tool to desktop users. The conversational interface lets traders and analysts query on‑chain data, generate smart alerts, and even prepare swaps, with manual confirmation...
South Korean Watchdog Expands AI Systems to Track Crypto Manipulation
South Korea’s Financial Supervisory Service upgraded its Virtual Assets Intelligence System for Trading Analysis (VISTA) with an automated detection algorithm that can identify price‑manipulation windows without human input. The new sliding‑window grid search scans every sub‑period in trading data, reproducing...
Tailoring CRISPR–Cas PAM Specificity via AI Models
Researchers have unveiled AI-driven models that accurately predict and redesign protospacer adjacent motif (PAM) sequences for a wide range of CRISPR‑Cas enzymes. The platform leverages deep learning on millions of natural and synthetic PAM datasets, achieving up to 40% reduction...
AI-Driven Thought Leadership Turns SME Insight Into Sales
Thought leadership should be a sales tool, feed search results, and build authority. That's a lot. But what if AI could help you surface and scale that internal expertise? By pairing AI tools with intentional workflows, you can capture unique...
Nano-vLLM: How a vLLM-Style Inference Engine Works
Nano‑vLLM is a minimal Python implementation that reproduces the core features of the widely‑used vLLM inference engine, including prefix caching, tensor parallelism, and CUDA‑graph optimizations. Benchmarks show it matches or slightly exceeds vLLM’s throughput despite its 1,200‑line codebase. The architecture...
Shield Raises $100 Million From Thrive Holdings to Accelerate the Growth of Its IT Services Platform
Shield Technology Partners announced a $100 million investment from Thrive Holdings to accelerate its AI‑first IT services platform. The funding will boost product innovation, expand the company’s national partner network, and fund strategic M&A. Shield’s AI‑driven tools, Sentinel and Spectre, have...
Google Ads Replacing Support Form With AI Agent
Google Ads has replaced its traditional support form with an AI‑driven chat agent, accessible via the usual support URL. The chatbot introduces itself as an AI advisor, prompting users to describe issues before offering assistance or escalating to human help....
DeepL Launches Voice API for Real-Time Speech Transcription and Translation Enabling Instant Multilingual Communication
DeepL announced the general availability of its Voice API, a cloud service that streams audio and delivers real‑time transcription plus translation into up to five target languages. The API is aimed at contact centers and BPO providers, allowing agents to...
Q&A: Why Deep Tech Investors Are Turning To The Secondary Market For Liquidity
Celesta Capital’s founding partner Sriram Viswanathan explains that abundant capital and a shrinking pool of high‑quality deep‑tech deals are fueling a surge in secondary‑market transactions. Investors are using secondaries to gain liquidity as private exits stretch, especially in AI‑driven hardware...
01C Launches Amara to Generate Full 3D Worlds From Simple Prompts
London‑based startup 01C has launched Amara, an AI‑powered platform that lets designers, game developers and 3D artists generate complete, production‑ready 3D environments from simple text or photo prompts. The tool is built natively for Unreal Engine and plugs directly into...
PDFgear Launches TextaVoice, a Truly Free Text-to-Speech Challenging Expensive TTS Subscriptions
PDFgear has launched TextaVoice.com, a completely free text‑to‑speech platform. The service converts up to 2,000 characters per request into natural‑sounding audio without sign‑up, ads, or usage caps. It offers over 30 languages, 236 voice styles, adjustable speed, pitch, emotion, and...
How SAP Is Modernising HMRC’s Tax Infrastructure with AI
HMRC has awarded SAP a contract to replace its legacy Enterprise Tax Management Platform with a cloud‑native solution built on SAP Business Technology Platform and RISE with SAP. The migration moves £800 billion of annual tax revenue processing to a UK...
SERPs with Ads + AI Overviews Grew by Over 394% in 2025
Google has dramatically increased ad presence on search results that feature AI Overviews, jumping from 5.17% in March 2025 to 25.56% by October—a 394% rise. The growth is uneven across placements, with bottom‑of‑SERP ads now appearing on roughly a quarter...
Former OpenAI Researcher Says Current AI Models Can't Learn From Mistakes, Calling It a Barrier to AGI
Jerry Tworek, a former OpenAI researcher behind the o1 and o3 reasoning models, argues that today’s AI systems cannot learn from their own mistakes, a flaw he sees as a major obstacle to artificial general intelligence. He left OpenAI to...
Controversial Proposal To Label Sections Of AI Generated Content via @Sejournal, @Martinibuster
A proposal by David E. Weekly suggests a new HTML attribute on the element to label AI‑generated sections of a page. It aims to meet the EU AI Act Article 50 requirement for machine‑readable AI content disclosure, addressing the gap...
Molecular Hamiltonian Learning Extracts Parameters From Stm-Iets Data for Single Molecules
Researchers at Aalto University introduced "molecular Hamiltonian learning," a machine‑learning framework that infers the full Hamiltonian of single‑molecule magnets directly from set‑point‑dependent scanning tunneling spectroscopy (STM‑IETS) data. By training on a library of theoretical spectra that include crystal‑field, Coulomb and...
Orbiit Recovery Ecosystem Advances the Use of AI and Machine Learning in Addiction Recovery
Orbiit Services Inc. unveiled an AI‑powered Recovery Ecosystem that interprets everyday mobile‑device interactions to monitor addiction recovery. The platform creates a behavior‑derived intelligence profile, generating real‑time Sober and Risk scores that clinicians can view on a unified dashboard. Interventions are...
From Algorithms to Advisors, Learn How Professionals Are Navigating Money in the Age of AI
AI is reshaping financial decision‑making by moving from pure automation to augmentation, but human judgment remains essential. Professionals are pairing AI‑driven insights—such as scenario modeling and stress‑testing—with personalized advice from financial planners, especially in high‑cost regions like California. This hybrid...
Norwegian AI Startup Watchdog Raises $600K to Catch Costly Invoice Errors
Oslo‑based AI startup Watchdog announced a $600,000 seed round led by venture studio Askeladden & Co and angel investors from OpenAI, Encube and Riff. The funding will accelerate development of its AI‑powered invoice verification tool that automatically cross‑checks invoices against...
Menotracker Wants to Fix Menopause Symptom Tracking — without Selling User Data
Menotracker, an AI‑powered menopause‑tracking app, has launched with privacy‑tech firm ConsentKeys to ensure user data is never stored or sold. The platform lets women log symptoms, connect wearables, and generate doctor‑ready reports while assigning pseudonymous identities that protect against breaches....
Richtech Robotics Partners with Microsoft to Advance Agentic AI in Real-World Robotics Applications
Richtech Robotics has teamed with Microsoft’s AI Co‑Innovation Labs to embed Azure‑powered agentic AI into its flagship Adam robot. The upgrade adds vision, voice and contextual reasoning, letting the robot factor in time of day, weather, promotions and equipment status....
Coder Launches AI Maturity Self-Assessment
Coder, a self‑hosted AI development platform, introduced a free AI Maturity Self‑Assessment and accompanying AI Maturity Curve. The tools let engineering teams benchmark AI adoption across development practices, operational controls, and governance readiness. By mapping questionnaire responses to the maturity...
Interview with Zijian Zhao: Labor Management in Transportation Gig Systems Through Reinforcement Learning
In this interview, Ph.D. candidate Zijian Zhao discusses his work on labor management in transportation gig platforms using reinforcement learning, covering order dispatch, pricing, and the challenges of large state and action spaces. He highlights novel MARL and single‑agent RL...
Ring Reunites Lost Dogs in New Super Bowl Spot
Ring, an Amazon subsidiary, debuted a 30‑second Super Bowl ad unveiling its new Search Party for Dogs feature. The tool leverages AI‑driven image recognition and opt‑in Ring cameras to coordinate neighborhood searches for missing pets. The spot, Ring’s first linear‑TV...
What Happens to the AI Exit Market if the FTC Cracks Down on ‘Acquihires’?
The FTC is signaling a shift toward treating both traditional acquihires and reverse‑acquihires as subject to the same antitrust standards that govern full mergers. Reverse‑acquihires—where a firm takes a minority stake or no money but hires key talent—have historically avoided...
ACTFORE Secures Patent for Template Identification and Matching Technology
ACTFORE announced the USPTO has granted its second patent for a Template Identification and Matching technology that fingerprints documents at the pixel level. The solution converts each page into image‑based signatures, enabling detection of structural similarities across massive, unstructured breach...
Apply Sports AI Tactics: Real‑Time, Personalized, Scenario‑Driven Business
4 practical AI lessons from sport. Sport is one of the best stress tests for AI, because decisions are fast, public, and high stakes. Here are 4 AI lessons every executive can steal from elite sport 👇 4) Fan Engagement...
Decentralized AI Networks Self‑Improve Exponentially, Outpacing Competition
A decentralized AI computer is an exponential learning machine. Each node can discover a way to improve itself. Test its theories. And report its results with each other. The whole network will improve together. Exponentially quickly. If you are going to compete with it...
Julie Guirado to Lead CHT as Executive Director
The Center for Humane Technology announced Julie Guirado’s promotion from COO to Executive Director, positioning her to lead the organization through a pivotal year for AI. Daniel Barcay, who guided CHT during the rise of generative AI, will transition to...
A Notable Development - Pegasystems Taps Into the Notes Replacement Market Opportunity
Pega announced Notes to Blueprint™, a specialized extension of its AI‑driven Blueprint platform aimed at modernizing legacy Lotus Notes applications. By leveraging assets from the recent Adopteq acquisition, the solution can analyze, extract, and re‑architect Notes databases into cloud‑native workflows....
Anthropic’s Claude Is Williams F1’s ‘Thinking Partner’ as AI Brands Race Into Sports
Anthropic has signed a multi‑year deal with the Atlassian Williams Formula 1 team, naming its Claude chatbot the “Official Thinking Partner.” Claude will be used by engineers and strategists for race strategy, car development, and operational decisions, with branding appearing on the...
These AI Notetaking Devices Can Help You Record and Transcribe Your Meetings
Physical AI notetaking devices are gaining traction as an alternative to software‑only transcription services, offering hands‑free recording for in‑person meetings. Products like Plaud Note, Mobvoi TicNote, Comulytic Note Pro and Anker Soundcore Work combine multiple microphones, on‑device AI and mobile apps...
NVIDIA AI Brings Nemotron-3-Nano-30B to NVFP4 with Quantization Aware Distillation (QAD) for Efficient Reasoning Inference
NVIDIA released Nemotron-3-Nano-30B-A3B-NVFP4, a 30‑billion‑parameter LLM quantized to 4‑bit NVFP4 while preserving BF16 accuracy. The model combines a hybrid Mamba2 Transformer Mixture‑of‑Experts architecture with a Quantization Aware Distillation (QAD) pipeline that replaces task loss with KL divergence to a frozen...
Agentic AI ROI Goes Beyond Revenue—Spot Real Wins
Why Agentic AI ROI Isn’t Just Revenue – Look Wider or Miss Real Wins Topics: ↳ Why Agentic AI ROI Isn’t Just Revenue – IBM’s Call to Look Wider or Miss Real Wins ↳ AI Tokens: Enterprise Spend Surges 20% – On-Prem Open...
IManage Delivers Trusted AI Search with Ask iManage
iManage has upgraded its Ask iManage AI assistant to answer natural‑language questions across the entire iManage Work platform, delivering structured, cited responses without pre‑selecting documents. The feature, now generally available at no extra charge, adds instant summaries, side‑by‑side comparisons and...
Biorce Raises $52M to Support Global Rollout of Its AI Clinical Trial Platform
Barcelona‑based health AI firm Biorce closed a $52 million Series A round, led by DST Global Partners and supported by existing backers Norrsken VC, YZR Capital and Mustard Seed Maze. The financing brings total capital raised above $60 million and will fund the...
Open-Source AI Pentesting Tools Are Getting Uncomfortably Good
Open‑source AI pentesting tools are reaching production‑grade performance, with BugTrace‑AI, Shannon, and the Cybersecurity AI Framework (CAI) each demonstrating distinct strengths in a lab test. BugTrace‑AI acts as a low‑noise reconnaissance assistant, flagging likely flaws without launching exploits. Shannon takes...
Match AI Capabilities to Tasks, Not Just Benchmarks
Choosing The Right AI In 2026 Is No Longer About Choosing The Right Model In 2026, choosing the right #AI comes down to matching #capability profiles to specific tasks, risk levels and business outcomes, rather than chasing benchmark winners. This...
CES 2026 Highlights Shift From Humanoid Hardware to Real-World Behavior, Says IntBot
CES 2026 demonstrated that robot hardware has reached a performance plateau, with machines easily executing complex movements like dancing and back‑flipping. The real challenge, according to IntBot, is now social intelligence—robots must understand context, people, and environments autonomously. IntBot highlighted its...
Leverage Claude as Your Smarter, Self‑Aware Problem‑Solving Clone
Are you a business person using Claude Code or Claude Cowork? Cool, grab my lazy hack. Rant, rant, rant, and rant some more. Describe your day. Describe your blockers. Share what you've tried. Give your criteria. Rattle off your resources....
Structured AI Rant Sessions Boost Productivity and Shipping
My most productive AI workflow is literally just complaining to Claude for 3 paragraphs then typing "help me fix this plz." 👉 Rant about your day and ask for help prioritizing. 👉 Rant about a project and ask for blindspots. 👉 Rant about...
AI Is Flooding IAM Systems with New Identities
The Cloud Security Alliance reports that AI‑generated identities are being treated like traditional non‑human accounts, exposing them to the same IAM weaknesses such as credential sprawl and unclear ownership. AI systems create and retire identities at unprecedented speed, overwhelming legacy...
State of Louisiana Partners with Persona AI on Humanoid Robotics Pilot at Steel Fabrication Plant
Louisiana’s economic development agency signed a memorandum of understanding with Persona AI to launch a humanoid‑robotics pilot at SSE Steel Fabrication’s plant in St. Bernard Parish. The program will collect real‑world data on human movement and task execution in a...
How to Build Memory-Driven AI Agents with Short-Term, Long-Term, and Episodic Memory
The tutorial presents a full‑stack memory engine that splits an AI agent’s context into short‑term working buffers, long‑term vector stores, and episodic traces. It leverages sentence‑transformer embeddings and a FAISS index to enable rapid semantic similarity search, while a policy...
OpenClaw Lets You Build Powerful AI Tools Fast
I spent 14 hours this weekend building with @openclaw. It did NOT disappoint. Those who say it’s overhyped are either scared, lazy, or both. Yes, there are security concerns to work around but like anything, there are tradeoffs. As long as you have...
AI Ready to File Taxes Without Human Help
Feels like the first year where AI is going to autonomously handle a lot of people's taxes. Either OpenClaw or Claude Cowork pointed at a folder of business docs, quickbooks reports, bank statements, docusign downloads, emails with accountant.
CFO’s Like AI But Don’t Trust It: Report
A Wakefield Research survey of 100 mid‑market CFOs shows 60‑77% plan AI adoption, yet a profound trust gap hinders execution. While 96% cite time‑saving as AI's top benefit, only 14% fully trust AI to produce accurate accounting data, and 97%...
A Coding and Experimental Analysis of Decentralized Federated Learning with Gossip Protocols and Differential Privacy
The tutorial implements both centralized FedAvg and a fully decentralized gossip-based federated learning system, adding client‑side differential privacy via calibrated Gaussian noise. Experiments on non‑IID MNIST data compare convergence speed, stability, and final accuracy across privacy budgets (epsilon values). Results...
Build Deployable AI Workflows at SaaStr AI 2026
Not just talks. At SaaStr AI 2026, we're running 100s of hands-on workshops where you'll build actual AI workflows for your company. Sales. CS. Product. Ops. Leave with something you can deploy Monday. May 12-14. SF Bay. https://t.co/IsMOos9H4H