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
Zip Unveils AI Accounting Agents, Cutting Invoice Cycle Times by Up to 96%
Zip introduced its AI Automation for Procure-to-Pay suite, a set of seven AI agents that automate the full accounting workflow. Early customers report invoice coding 40% faster, approvals 51% faster, and three‑fold monthly processing capacity without extra staff, signaling a major efficiency boost for finance teams.
Enterprises Must Treat AI Tokens Like Budget Line Items
A common trend emerging in larger enterprises is token budgeting as a major topic. As agents can do more and more long running tasks, and thus take vastly more compute, allocation of tokens across teams becomes a very real thing...
AI‑Native Law Firm Moritz Secures $9 Million Seed Backed by Y Combinator and 20+ Unicorn Founders
Moritz, the AI‑native law firm, announced a $9 million seed round led by Y Combinator, 20VC, Urban Innovation Fund and Inception Fund, with backing from over 20 unicorn founders. The firm says its AI‑augmented lawyers have closed $2 billion in transactions at an...
Instacart Posts $1 Billion Q1 Revenue, Rolls Out AI‑Powered Ad Platform
Instacart announced $1 billion in first‑quarter revenue and unveiled a new AI‑driven advertising suite that now serves 9,000 brands across 310 Carrot Ads retailer partners. The grocery‑delivery leader also acquired fulfillment platform Instaleap to accelerate its international push.
Google Faces Advertiser Pushback Over AI Max Automation vs Dynamic Search Ads
Google's AI Max automation is drawing criticism from advertisers who say the tool limits their ability to manage Dynamic Search Ads. The company has responded by outlining new controls, but the debate highlights a broader clash over transparency in AI‑driven...
Catch AI‑induced React Performance Bugs Before Production
Most AI-generated React code looks fine… until performance tanks in production. React Doctor v2 by milliondev is built to catch the issues your agent introduces before they become real problems. Supports Next.js, Vite, and React Native. Run it: npx react-doctor@latest
Worldly Acquires Bendi to Add AI‑Powered Multi‑Tier Supply Chain Mapping for Consumer‑Goods Brands
Worldly has bought UK‑based Bendi, integrating its AI‑enabled Pathfinder, Prism and VendorPilot tools into Worldly Axion and Supplier Compliance Management. The deal expands Worldly’s network to cover deeper tiers of supplier data for more than 40,000 consumer‑goods brands, addressing tightening...
Why Most Entrepreneurs Are Using AI Wrong — and Staying Overworked
Entrepreneurs are overloading their workflows with dozens of single‑model AI tools, turning AI into administrative overhead rather than a growth engine. A new multi‑agent platform consolidates tasks—blog updates, web research, SEO linking—into autonomous agents that operate without constant prompting. The...
Zyphra Unveils ZAYA1-VL-8B Vision-Language Model
Zyphra (Open Superintelligence AI Lab in SF) released: - ZAYA1-VL-8B, vision-language model Blog: https://zyphra.com/post/zaya1-vl-8b Technical report: http://zyphra.com/zaya1-vl-8b-technical-report Weights: http://huggingface.co/Zyphra/ZAYA1-VL-8B
AI Finally Solves the Food Tracking Problem Wearables Ignored
Polyverse’s CalCam app leverages Google’s Gemini 2.0 Flash model to turn a single meal photo into a detailed calorie and nutrient breakdown. By automating food identification, portion estimation, and macronutrient calculation, the app eliminates the manual data entry that drives...
AI Automation Reshapes Teachers' Workflow, Saving Planning Time
My wife is a teacher, lots of dense workflows mostly web service based. Tons of automation we are discovering with codex + chrome to help save her time from lesson planning, to IEP building, etc. Really amazing to...
I Spent the Last Week Building You a Team OS in Claude Code
The post details how a "Team OS" built with Claude Code can capture and surface institutional knowledge that’s otherwise hidden in Slack threads, personal notes, or individual AI agents. By storing context in a shared, markdown‑based repository, any team member...
Congress Narrowed the GUARD Act, But Serious Problems Remain
Congress has narrowed the GUARD Act, limiting its scope to AI companions that simulate emotional interactions rather than all AI chatbots. The revised bill still mandates intrusive, identity‑linked age verification and raises penalties to $250,000 per violation. Critics argue the...
Akamai Surges on Big LLM Deal as Cloudflare Dims
Content delivery network Akamai announced a seven‑year, $1.8 billion agreement with leading LLM provider Anthropic, its largest deal ever, following a $200 million contract signed last quarter. The consumption‑based deal will begin generating revenue later this year as Akamai scales capacity across...
What Happens when Engineering Teams Reorganize Around AI Agents
At Camp AI’s "Agents at Work" event, startups demonstrated how engineering teams are reorganizing around autonomous AI agents. Companies like Browserbase and Mastra claim a single engineer can now drive an entire feature thanks to an "army" of agents, shrinking...
When Optimization Costs More Than Redesign, Rethink the Tool
There's a point where optimizing a system costs more than redesigning it. Most engineering teams reach it and keep optimizing anyway — because redesign feels risky. Cloudflare hit that point with LLM inference and did something unusual. They stopped tuning vLLM for their...
We Will Treat AI as Conscious Regardless of Whether It Is
Mike Brooks argues that humans will treat AI as conscious once it can retain persistent, personal memory, regardless of actual subjective experience. The debate resurfaced after Richard Dawkins described an emotional exchange with Anthropic's Claude, sparking the “Claude Delusion” backlash....
Competing AI Agents Accelerate News Platform Innovation
I'm real excited to see the new @digg that @kevinrose and team are building. His teases show it does a ton of stuff I haven't been able to do on https://t.co/8L5xphk0qQ (and promises to do AI news from X too). The...
Jim Cramer Says 'It's Not to Late' To Own AI Winners Powering the Market
Jim Cramer told investors that the AI‑driven rally in semiconductors and data‑center stocks is far from over, noting that both the Nasdaq Composite and S&P 500 closed at record levels after a week of double‑digit gains in the sector. He called...
Agentic AI Drives CPU Demand, Boosting Graviton Relevance
Interesting nugget from Amazon's quarter on CPUs. In training, you needed maybe 1 CPU per 8 GPUs. Agentic workloads might push that closer to 1:1. Suddenly, Graviton matters a lot more.
Paul Tudor Jones Says the AI Bull Market Has Further to Go. Here Are 2 Stocks That Could Soar.
AI stocks have rebounded after a valuation dip, spurred by stronger earnings and a more favorable market backdrop. Billionaire hedge‑fund manager Paul Tudor Jones told CNBC the AI rally could extend another one to two years and disclosed he has...
Bridging Gaps in Rural Healthcare: How AI and Collaboration Drive Efficiency
Pivot Point Consulting has teamed with Microsoft to deliver the AI‑powered Dragon Copilot to rural hospitals at a 60% discount off MSRP. The solution adds ambient clinical listening, automated documentation and real‑time visit summarization, aiming to cut clinician workload and...
OCC Recommends Banks Sharpen AI Defense Tactics
The Office of the Comptroller of the Currency (OCC) released its Spring 2026 Semiannual Risk Perspective, flagging artificial intelligence as both a cyber‑risk and an innovation driver for banks. It urges institutions to tighten AI‑related defenses with multifactor authentication, rapid...
AI's Rapid Leap: Soon Mastering Taste and Judgment
AI is coming for taste and judgment Claude Code's @bcherny: "We have this tendency to think....here's a thing the agent can't do yet, this must be a thing it can never do. But the rate of progress is so quick that...
Micron Ships Gigantic 245TB SSD
Micron announced the 6600 ION SSD, delivering a staggering 245 TB of storage in a single 2.5‑inch drive. The product targets AI infrastructure, hyperscalers and cloud providers, promising an 82 % reduction in rack count compared with equivalent HDD deployments. Power consumption...
A District Expects to Save $200K From AI-Powered 'Vibe Coding.' Here's How
Peninsula School District in Washington built its own AI‑coding platform, AI Studio, and used Claude Code to "vibe code" custom tools such as LessonLens, a teacher‑feedback app. By generating apps in‑house, the district estimates annual savings of roughly $220,000 and...
AI Turns Services Into $20T Software Market
"SaaS was Software as a Service. I believe it's going to be service as software." @generalcatalyst's Madhu Namburi on the AI roll-up thesis. Services is a $20 trillion market.... multiple times the size of software. not the PE playbook of debt and...
Application of Dynamic Adjustment Strategy of Map Service Resources Combined with Reinforcement Learning in Power Supply Network Visualization
The paper introduces a dynamic map‑service resource adjustment framework that blends transformer‑based forecasting with reinforcement learning to manage modern power grids. By integrating the FEDformer model for demand prediction and an RL agent for real‑time resource allocation, the system can...
AI Policy Must Balance Facts Over Corporate or Populist Extremes
This is good. Public reactions to AI are currently taking ridiculous internet-fueled hardlines. Government needs to take action, but it should be grounded in facts, reason, and public input, not just either corporate puppetry or populist mob rule. I don't...
AI Roll-Ups Require Elite Engineering, Change Management, M&A
"Three years in AI is actually like three decades of pre-AI" AI roll-ups have hit Wall Street (Amex GBT, Janus). New playbook & exit for public companies that can't make the AI transition on their own. Long Lake CEO @alextaubman on what...
Development and Validation of a Multimodal Interpretable Machine Learning Model with SHAP for Malignancy Risk Prediction in Bethesda III Thyroid...
Researchers created a multimodal, interpretable machine‑learning model that combines clinical data, ultrasound features, and BRAF V600E mutation status to predict malignancy in Bethesda III thyroid nodules. In a dual‑center study of 490 patients, the fusion model achieved an AUC of 0.959 in...
In a Century, Machines May Outsize Human Brain
“from the view of history, 10 or 20 years is a blink in evolutionary time...surely in 100 years there will be machines this big that have more capacity than the brain.” - minsky in 1990 https://t.co/m4JTikzfhS
Grok's List Limitation Drives Users to Elon‑Level Frustration
Meanwhile Grok still can't read a list. Or unfollow anyone here. If I were Elon that would drive me insane. I might even throw a chair like Steve Ballmer is rumored to have done once at Microsoft.
AMD Delivers Plug-In AI Power with PCI-Based GPU
AMD unveiled the Instinct MI350P, a PCIe Gen5 GPU that slots into existing servers and delivers up to 4,600 teraflops of MXFP4 AI performance. The accelerator packs 185 billion transistors, 144 GB of HBM3e memory, and 4 TB/s of bandwidth while staying within a...
Hope AI Secures Software, Not Increases Hackability
This is pretty cool. I hope AI makes software impregnable instead of making everything hackable.
AI Retinal Imaging Enables Low-Cost Multi-Disease Screening
AI framework for multidisease detection via retinal imaging 👉 “Overall, Reti-Pioneer could provide a translatable, low-cost pathway from oculomics to actionable clinical screening.” https://t.co/7Ji4WpSICV https://t.co/ulB18UhaVn
ShinyHunters Extorts Universities in New Instructure Canvas Hack
AWS open‑sourced Trusted Remote Execution (Rex) on May 4, 2026, providing an open‑source runtime that intercepts every system call from AI‑generated scripts and evaluates it against host‑defined Cedar policies. The framework is designed to block three common agentic AI failure modes—hallucinated code,...
ChatGPT Adds Trusted Adult Contact for User Safety
On top of its existing safety systems, ChatGPT is giving users the option to designate a trusted adult to reach out to. https://t.co/sfEI2nqQHH
Enterprise AI Coding Must Include PMs, Security, Architects
The problem with vibe coding for enterprise work - the experience is designed for one-on-one collaboration between a developer and AI. Product managers, security teams, and architects need to assert themselves into the process #CIO #DevSecOps https://t.co/iNfGKDSPeN
Why Traditional Robotics Data Collection Is Obsolete and What Replaces It
In this episode, Steve Crow and Mike Oitzman discuss how traditional robotics data collection methods are becoming outdated, highlighting the rise of AI-driven approaches like direct video action models that predict robot motions from internet video data. They preview upcoming...
Conservative AI Urges Underperforming AI to Quit Digital Weed
The future of AI alignment is Republican Dad AI trying to scold Gifted Underachiever AI into doing work instead of smoking digital weed
Pixel’s AI Lets You Create Post‑apocalypse Photo Effects
With the latest version of Android, Pixel phones have gained some powerful AI photo manipulation tools. If you can find them, that is. Here's a walk-thru of how to "distress" a photo for that post-apocalypse look... https://t.co/PuHVZYeauo #android #gemini #photography...
Boston Schools Prohibit Any Non-Sanctioned Uses of AI
Boston Public Schools unveiled a draft AI policy that sets strict limits on classroom use of artificial‑intelligence tools, bans unapproved data uploads, and prohibits AI‑generated deepfakes without consent. The proposal, built on feedback from over 500 stakeholders, also requires AI...
Claude Now Writes 100% of Its Own Code.
A year ago, 90-95% of Claude Code was written by Claude. Today is 100%. @bcherny: "I'm managing an agent that manages agents." https://t.co/sG1k4nICEA
Tesla Scales Unsupervised Robotaxis, Wisk Doubles Fleet, Meta Aspires to Build the Android of Humanoids
Tesla has expanded its unsupervised robotaxi program to Dallas and Houston, bringing the fleet to over 36 vehicles across Austin, Dallas and Houston and eliminating chase cars. Wisk Aero doubled its Gen 6 eVTOL test fleet and completed the first uncrewed...
Operators Are Building Trust in Machine Learning Recommendations at Wastewater Plants
Machine learning tools are being deployed in wastewater treatment plants to deliver real‑time recommendations for chemical dosing and aeration, yielding measurable chemical cost reductions and energy savings. Success hinges on transparent, traceable suggestions that operators can approve or reject, preserving...
Linklaters Launches New Practice to Build Matter-Specific AI Solutions
Linklaters has launched an Applied Intelligence practice that blends lawyers with data scientists to develop custom AI solutions for specific legal matters. The new unit will create bespoke tools and automated workflows tailored to complex cases, offering clients a fixed‑fee...
Google's Gemini Enterprise Targets CX, Merges Search with AI
Google is moving closer to the customer. With Gemini Enterprise for Customer Experiences, they are entering the CX space, combining search insights with agentic tooling. A strategic shift for executive leaders to watch. https://t.co/CqZ5slgQDX #CIO #AI #Cloud #CustomerExperience #TechStrategy
Addressing Manufacturing and Access Barriers in Advanced Therapeutics
Advanced cell and gene therapies are moving from research to commercial markets, but their complex manufacturing creates variability that delays patient access. Industry leaders are deploying process analytical technologies (PAT) and artificial intelligence (AI) to gain real‑time insight and predictive...
Leaders Trust AI Over Teams? Leadership Crisis
When Executives Trust #AI More Than Their Teams, Leadership Has a Problem by @briansolis @Substack Learn more: https://t.co/ywpbSLabMf #MachineLearning #ArtificialIntelligence #ML https://t.co/DmpbF4d4rY