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
Many Smartphones Don’t Detect Face Biometrics Spoofs or Properly Warn Consumers
Which? tested 208 Android smartphones and found 64% vulnerable to 2D printed photo spoofs. iPhones remain largely immune due to depth‑sensing Face ID, while some newer Pixel models also resisted flat images. Many Android manufacturers, including OnePlus and Motorola, fail to warn users that facial unlock is a convenience feature, and Samsung’s older Galaxy models lack adequate presentation‑attack detection. The report urges users to rely on PINs or fingerprint sensors for higher security.
Tracking Coding Agents' Tool Usage with 10k Queries
Most of our usage is now coming from agents so to monitor how they use and mention our tools, with @DAKlingbeil, we're now running regular 10k queries to all the most popular coding agents, share the traces on @huggingface and...
3 Data Trends Shaping the Race to AI Across Industries in 2026
Snowflake’s Data Trends 2026 report identifies three cross‑industry AI imperatives: agentic AI is moving from pilot to production, the data foundation remains the chief bottleneck, and governance with semantic standards is emerging as a competitive moat. The report cites that...
MODEX 2026: All AI, All the Time
At MODEX 2026, AI topped the MHI Annual Industry Report as the most disruptive technology for supply‑chain and logistics, with 70% of surveyed professionals expecting major impact within the next decade. Companies such as Carvana and Disney highlighted AI‑driven analytics...
Data Poisoning: Yet Another AI Threat: Artificial Intelligence Trends
Data poisoning, once viewed solely as a security attack, is now being adopted by companies as a defensive tool to control how AI models ingest their content. By inserting subtle errors, honeytokens, adversarial perturbations, or structural noise, firms can create...
AI Giants Pivot to Biology for Next Value Wave
if you've been paying attention, you know there's been a serious vibe shift in AI this month 3 major AI labs (anthropic, openai, amazon) all just went hard into biology. they're literally telling you this is where the next wave of ai...
Guest Post- AI Refusal in Legal Research Instruction – Part 1: Introduction
Ben Doherty, head of library instruction at the University of Virginia Law School, proposes an AI‑refusal policy for his Advanced Legal Research (ALR) course beginning Spring 2027. He argues for a return to analog, process‑based teaching methods—poster‑board charts, handwritten work,...
Claude Skill: Instructions Audit
Claude’s new Cowork audit skill scans a user’s CLAUDE.md file and the actual workspace to evaluate instruction quality. It assigns each section a rating—Strong, Adequate, Thin, Missing or Outdated—and flags undocumented folders, ghost directories, and unconnected tools. The skill then...
AI Payoff Remains Distant as Firms Keep Spending, PwC Finds
PwC’s latest executive survey finds 81% of C‑suite leaders believe meaningful AI returns are at least a year away, even as 74% plan to maintain or increase AI spending over the next 12 months. The study places CFOs at the...
AI Tools Free Time to Focus on Product
Spent more time with AI assisted development this weekend. I spend less time thinking about code and more time thinking about the product.
ZoomInfo Adds Free ChatGPT Connector with Enrichment Tools
@ZoomInfo is expanding our partnership with @OpenAI with our @ChatGPTapp connector. Open to ANYONE - free sign up gets you 100 free credits. We've added Intent, Contact and Company Enrichment, Find Similar Companies, and Account Context that marries your...
UCSD: AI-Enhanced Microscopy Produces Crisp, Real-Time Video Inside Live Cells
UC San Diego engineers have unveiled an AI‑driven upgrade to structured illumination microscopy called unrolled blind‑SIM (UBSIM). The algorithm delivers images twice as sharp as conventional microscopes and streams video at up to 50 frames per second. By embedding optical physics...
Apple May Have Hidden a Big Clue About the Future of Siri in This Graphic
Apple’s WWDC 2026 teaser graphic features a glowing “26,” which Bloomberg interprets as a hint that Siri will receive a major visual overhaul in iOS 27. The update is expected to integrate Siri into the Dynamic Island, expanding it with a glowing...
The Two Dimensions of Agent Skills
The post proposes a two‑axis model for defining agent "skills": what the skill describes (persona, tool, or workflow) and how it executes (stateless or stateful). It recommends adding explicit `type` and `execution` fields to a skill’s metadata file to make...
AI and CMMC: A Double-Edge Sword for Defense Contractors
The Pentagon’s CMMC program forces defense contractors to safeguard controlled unclassified information, and the surge of generative AI has complicated compliance by expanding assessment boundaries and introducing new attack vectors. Employees may inadvertently feed CUI into commercial large‑language models, risking...
Accelsius Announces General Availability of NeuCool IR150 and Launches NeuCool HyperStart Program
Accelsius announced the general availability of the NeuCool IR150, the first fully integrated 42U rack that combines a two‑phase coolant distribution unit with up to 150 kW of cooling capacity. The plug‑and‑play system is designed for hyperscale, neocloud and enterprise data...
Fintech Apps and AI Tools Are Transforming Money Management : Research
Plaid’s Spring 2026 research shows AI has moved from a novelty to a core expectation in personal finance, with 55 percent of Americans using AI for money‑related tasks in the past year. Users report high satisfaction—86 percent say AI clarifies their financial health,...
OpenAI's Codex Now Watches Your Screen to Remember What You're Working On
OpenAI introduced Chronicle, a new memory layer for its Codex app that records screen activity and turns it into local Markdown summaries. The feature runs in the background on macOS, storing recordings for up to six hours before deletion. Currently...
Bootstrapped Freepik Turns AI Threat Into $1B Success
2020: Freepik bootstrapped to €250M. 2 years later: AI nearly killed them. Here's how they made a $1B comeback 👇 Three co-founders already had exits under their belt when they started Freepik in 2010. Joaquín Cuenca (Loquo to eBay ,...
Kimi Vendor Verifier – Verify Accuracy of Inference Providers
Moonshot’s Kimi team released the Kimi Vendor Verifier (KVV) alongside the K2.6 model to let users confirm that third‑party inference services reproduce official benchmark scores. The open‑source tool runs six targeted tests—including OCR, vision preprocessing, long‑output stress, tool‑call consistency, and...
ISED Inquired About ‘Opt-Out Mechanism’ for AI-Generated News Summaries on Tech Platforms
Innovation, Science and Economic Development Canada (ISED) has asked what a fair opt‑out mechanism would look like for AI‑generated news summaries on platforms such as Google’s AI Overviews. News publishers argue these summaries siphon traffic, with up to 60%...
85% of Employees Say AI Training Doesn’t Help Them Do Their Jobs
Docebo’s 2026 AI Readiness Gap study finds 85% of employees say training doesn’t help them use AI at work. Gartner forecasts 20% of firms will cut >50% of managers by 2026. The U.S. Department of Labor allocated $85 million for apprenticeship...
Cloudflare Forces AI Crawlers to Follow Canonical URLs
This seems to be flying under the radar -> Redirects for AI Training enforces canonical content Once a site opts in via Cloudflare: "When a request arrives from a verified AI Crawler, Cloudflare reads the response HTML. If a non-self-referencing canonical tag...
AI Robotic Dogs Accelerate Disaster Response and Safety
#AI #Robotic Dogs Lead the Way in Fire and Disaster Response, Enhancing Safety and Speed by @IntEngineering #Robotics #Engineering #Innovation #Technology #TechForGood https://t.co/ZFuzKA31af
Reading Today's Open-Closed Performance Gap
The blog argues that the open‑vs‑closed LLM performance gap cannot be reduced to a single number, as benchmark suites like the Artificial Analysis Intelligence Index hide nuanced capability differences. Closed, frontier labs are pouring massive funds into reinforcement‑learning‑with‑verifiable‑rewards (RLVR) environments...
Anthropic's Philosopher Explores Claude's Personality and Consciousness
NEW: on the @NewcomerMedia podcast, Anthropic's philosopher queen @AmandaAskell. Meet the person charged with developing Claude's personality and ethical core. I ask whether Claude experiences consciousness. She's not ruling it out. https://t.co/hcgfnhhjBE
Companies Overestimate Consumer Comfort with AI Agents
Businesses are overestimating how customers are comfortable with using AI agents. (Blue bar / business is higher than grey bar /consumer on every question) #AdobeSummit #AI #cmo https://t.co/acbhpkA281
Google Cloud’s NEXT Big Moment
Google Cloud is leveraging rapid AI advances, especially its Gemini model, to accelerate growth, reporting Q4 cloud revenue up 48% to $17.7 billion and a backlog projected at $240 billion by year‑end. The upcoming Cloud Next event in Las Vegas will showcase...
AI's First Use Turns Skeptical Radiologists Into Advocates
In healthcare, you only understand how game-changing AI is once you use it. What strikes me most is how reliably opinions flip. The skeptic radiologist goes from “I don’t need this” to “I can’t imagine working without it.” Not occasionally. Almost every time.
Gemini Launches MCP API for AI Trading Bots
Calling all robots 🤖. Check out @Gemini's new MCP API, designed specifically for AI agents to trade predictions, crypto, and more.
How Luxury Brands Are Quietly Leaning Into Artificial Intelligence
Luxury brands are quietly integrating artificial intelligence across operations, customer experience, and creative design. The AI market serving luxury is forecast to rise from $1.2 billion in 2024 to $5.6 billion by 2034, growing over 16% annually, with 45% of brands already...
Adobe Shifts From Managing to Orchestrating Experiences with Agentic AI
.@Adobe Doubles Down on Agentic AI — But the Hard Work May Be the Operating Model https://t.co/XBcswHmUhn @cmswire @DomNicastro "They have lost the ridiculous language of 'managing' experiences and adopted the stance that experiences are orchestrated," - @lizkmiller @constellationr @AdobeSummit
Funding Boosts AI for 3D Protein and RNA Mapping
Funding will support the development of advanced AI methods to map proteins and RNA structures in three dimensions. https://t.co/L6ZSpg8iMW #infrastructure #IoT #AI #5G #cloud #edge #futureofwork @virginia_tech https://t.co/7K7X1E9NpH
Takes on Automating Alignment
Recent AI models have shown a knack for tackling long‑horizon tasks when a clear metric guides progress, as demonstrated by MirrorCode’s ability to generate tens of thousands of code lines using extensive test suites. Anthropic’s Automated Weak‑to‑Strong Researcher further proved...
Meta Pours $200B Into AI, Trains Fiber Technicians
Meta has committed ~$200B on AI Capex. There’s such a shortage of fiber technicians, it launched a free training program with CBRE. Beginning to think Meta is behind that “Data Centre” game on Steam where you buy racks, set up servers...
Turn One Prompt Into a Ready Google Slides Deck
🖥️ Rough lesson outline plus one prompt in Gemini equals a full slide deck. 📊 It exports straight to Google Slides ready for you to customize. 🎯 This one has been a super popular strategy from my recent workshops. https://t.co/yYaO02BkT3
Early Adopters: 42% Rely on AI for Purchases
The 25% using AI for purchasing decisions jumps to 42% for early adopters. #AdobeSummit https://t.co/Xgwrf83UM1
Fed Nominee Warsh Calls AI an Unmatched Productivity Wave
JUST IN: Nominated Fed Chair, Kevin Warsh calls AI "the most productivity-enhancing wave of our lifetimes—past, present and future."
Youth Jobs Hit in AI Fields, Not Solely AI
Fascinating paper. To me, it simultaneously strengthens the thesis that something is affecting youth employment in AI-exposed industries, while slightly *weakening* the argument that it's AI, at least conventional AI displacement effects in the way we've been talking about them.
Local GPU App Turns 2D Images Into 3D Meshes
Someone literally just made a desktop app that transforms 2D images into 3D models, running 100% LOCALLY on your own GPU 🤯 Meet Modly. Zero cloud, zero crazy API fees. Just drag, drop, and BOOM: instant 3D mesh 💥 Fully open-source. https://t.co/eDKn1U8IhQ
AI May Never Achieve True Consciousness, Says Skeptic
"Why AI will never be conscious" my new conversation with the exceptional @CosmicSkeptic. The 'never' in the title is a bit strong - e.g. it depends what we mean by AI - but that's why you should listen to the...
Claude Claims Native German Translation, Skipping Extra Steps
I asked Claude to convert a German-language PDF to plain text "so I can send it to Google Translate" and at the end he was like, "FOR THE RECORD, I can speak and translate German perfectly well myself"
Log Mistakes and Add Corrections to Cut Token Burn
I spoke about logging mistakes (and reducing token burn) in this video. As for mistakes it’s important to include corrections to ensure you don’t reinforce the mistakes.
Track AI/LLM Traffic and Conversions with Looker Studio Template
Time to track AI / LLM traffic and conversions. Clone this Looker Studio Template. Link -> https://t.co/nGXnF2Duut
First Day with Claude Could Spike Labor Productivity
Would not surprise me to see a companion piece to this in the near future to the effect of “What happens to labor productivity and the economy when it’s always someone’s first day using Claude” https://t.co/3mHD18wYZ3 bcc: @MattZeitlin https://t.co/uHQUAJLFtq
Rising AI Tool Fees Threaten Small AI-Only Businesses
Oh man, this is definitely gonna happen: - someone builds biz on AI tool - AI tool increases fees - aforementioned biz no longer profitable God… I hadn’t even thought of this. Feels like sometime in 2027, this’ll happen and crippled small AI-only businesses.
Learn AI Fast by Watching Pair Prompting in Action
The fastest way to get good at AI is to watch someone good use it. Pair prompting makes that possible in public, or semi-public, channels where the work accumulates and others learn by following along. This essay is my attempt to name...
New Google TPUs Arrive, Compute Shortage Trumps GPU Debate
$GOOGL will launch its next generation TPUs this week. The narrative in the press will be Google vs. $NVDA. I strongly suggest not spinning wheels on the TPU versus GPU debate. Right now all capacity is good capacity because we don't...
AI Fuels CPU Renaissance, Lifting AMD, Arm, Intel
AI-Driven 'CPU Renaissance' To Boost AMD, Arm And Intel, Analyst Says https://t.co/QTANhz2sSt $AMD $ARM $INTC $RMBS
Humanoid Robot Beats Human Marathon Record
Nothing to worry about, folks. -The Future A year ago in Beijing’s humanoid-robot half-marathon race, the first runner to cross the finish line took more than 2½ hours. In this year’s event, the champion beat the fastest human ever. https://t.co/1Sel66fSCa via...