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
Automation Is Real—Mass Layoffs Aren’t #152b
The anticipated AI‑driven job apocalypse has not materialized. Vanguard’s analysis of roughly 140 occupations highly exposed to AI shows those roles grew 1.7% annually from 2023 to 2025, outpacing the 0.8% growth of other jobs, while wages rose 3.8% versus 0.7% elsewhere. Parallel research from McKinsey, the ILO and the OECD confirms that automation is reshaping tasks rather than wiping out entire occupations. The data suggests AI is augmenting work, not eliminating it.
Codex Xhigh Cracks 6‑month Bug by Digging Into GTK4
Ahhhh, Codex 5.3 (xhigh) with a vague prompt just solved a bug that I and others have been struggling to fix for over 6 months. Other reasoning levels with Codex failed, Opus 4.6 failed. Cost $4.14 and 45 minutes. Full...
Top AI Founders Pitching Opportunities at Conviction NYC
3.11 in NYC: @Conviction’s “Mixture of Experts” ✨a curated set of the founders you need to know in AI, pitching you (engineer, researcher, product person) their company to join 💫 https://t.co/hXUsZctqfH
New HEPI Report: ‘Being Indispensable: Capabilities for a Human-AI World, the ‘FUTURES’ Framework’
The Higher Education Policy Institute (HEPI) released Report 198, titled “Being indispensable: Capabilities for a human‑AI world, the ‘FUTURES’ framework,” urging universities to move beyond cautious AI pilots and embed generative AI into curricula. The report stresses that while GenAI can...
Chef Uses Hillclimbing, Sets SOTA on FoodBench
at a top agent lab even the chef is hillclimbing imo this is SOTA for @MatthewBerman 's foodbench https://t.co/piai2q5HVc
Why the Future of AI Depends on a Portable, Open PyTorch Ecosystem
Red Hat is championing an open, portable PyTorch ecosystem to ensure AI models run on any accelerator, cloud, or hardware. By contributing to projects like vLLM, vLLM‑CPU, OpenReg, and advanced kernels, Red Hat aims to democratize inference and reduce reliance...
Introducing Modular Diffusers - Composable Building Blocks for Diffusion Pipelines
Modular Diffusers launches a composable framework that breaks diffusion pipelines into interchangeable blocks such as text encoding, denoising, and decoding. Developers can assemble, replace, or run individual blocks, enabling lazy loading, memory‑efficient inference, and easy experimentation with models like FLUX.2‑Klein 4B....
Data Center Intelligence at the Price of a Laptop
Alibaba’s Qwen3.5-9B open‑source model can run on a standard 12 GB RAM laptop, delivering performance comparable to Claude Opus 4.1. The author’s token‑usage analysis shows that a $5,000 laptop pays for itself after processing roughly 556 million tokens—about a month of typical workload—making...
John Wiley & Sons Inc (WLY) Q3 2026 Earnings Call Transcript
John Wiley & Sons reported FY2025 adjusted revenue up 3% as research, academic and AI licensing drove growth, offsetting weakness in professional channels. AI licensing contributed $40 million, highlighted by an $18 million Q4 agreement, while free cash flow rose 10% to...
MCP Security: Implementing Robust Authentication and Authorization
The Model Context Protocol (MCP) is becoming a core interface for AI agents to invoke tools and access enterprise data. To mitigate rising security threats, the latest guidance recommends treating MCP servers as OAuth 2.1 resource servers and adopting modern authentication...
Now in Public Beta: Store Terraform State in Pulumi Cloud
Pulumi Cloud has launched a public‑beta Terraform state backend, letting teams point their Terraform or OpenTofu CLI at Pulumi Cloud without altering HCL. The service stores state encrypted, provides automatic locking, versioned history, RBAC and audit policies, and surfaces the...
AI Accelerates Vulnerability Fixes, Sparking Dual Optimism
Being at UnpromptedAI conference in SF this week seeing extent of using AI for vulnerability discovery and for accelerating fixes has me want to make a slight correction to this post: 1. I am short term *very* pessimistic 2. I am *even...
Systems Design and the Semantic Revolution
The article revisits Postel’s Law—strict sending, generous receiving—as the foundation of internet interoperability and argues it now underpins the semantic revolution driven by large language models (LLMs). After costly cycles of syntax‑focused standards like XML and blockchain, LLMs are tackling...
‘AI, AI And AI’: CommonSpirit Health at Home CEO On The Forces Shaping 2026
CommonSpirit Health at Home is charting a 2026 roadmap that moves care upstream by using EHR‑based eligibility triggers to engage patients earlier in hospice and home‑health pathways. The provider plans a measured AI rollout focused on predictive analytics and operational...
Open Hidden Open Thread 423.5
OpenAI has signed a new contract with the U.S. Department of Defense, expanding its involvement in Pentagon projects. The agreement’s surveillance language contains numerous ambiguities that could allow broad data collection. Critics on LessWrong highlight potential loopholes that may undermine...
MicroCloud Hologram Advances Deployable Quantum Recurrent Neural Network Technology
MicroCloud Hologram Inc. (HOLO) announced a Quantum Recurrent Neural Network (QRNN) built around a novel Quantum Recurrent Block (QRB) architecture designed for noisy intermediate‑scale quantum (NISQ) devices. The QRB acts as a modular, repeatable subcircuit that drastically reduces coherent‑time consumption,...
Nvidia Hiring for Orbital Data Center System Architect, as Space Compute Market Grows
Nvidia announced a senior hire for an orbital data‑center system architect, offering a base salary between $224,000 and $356,500. The role will design end‑to‑end AI compute solutions that operate from the GPU chip through satellite platforms and inter‑satellite links. The...
Human Cringe Beats AI Polish in Pitch Authenticity
I can always tell when someone used AI to write their pitch. Not because it’s bad — it’s usually polished. It’s because it has no embarrassing, specific, human details in it. The weird personal stuff IS the good stuff. AI...
Scale Ambient AI with Strategic Partnerships
Healthcare technology firms are increasingly embedding Ambient Clinical Intelligence (ACI) into EHR, telehealth, and care‑management platforms to cut administrative load and speed revenue cycles. Building ACI models internally demands extensive engineering resources, prompting many vendors to seek external expertise. Suki’s...
MWC 2026 Launches AI‑Driven 5G‑to‑6G Era
Industry Analyst: MWC 2026 ushers in new AI - 5G - 6G Era to wireless world https://t.co/kuBOGueI89 @jeffkagan #jeffkagan #industryanalyst #ictindustryanalyst #ai #aianalyst #aiindustryanalyst #analyst #analysts #wirelessanalyst #telecomanalyst #wirelessindustryanalyst #telecomindustryanalyst #tech #technology #techanalyst #technologyanalyst #techindustryanalyst #pr #publicrelations #pressrelease #marketing #ceo #cmo #caio...
Report: 5 AI Agent Trends to Watch in Healthcare & Life Sciences
Google Cloud’s 2026 report spotlights five AI‑agent trends reshaping healthcare and life sciences, from automating clinical paperwork to accelerating drug discovery. The analysis stresses that unlocking AI value requires leaders to discard legacy mindsets and foster a culture that embraces...
The Big Questions That Will Decide the Consumer AI War
The episode examines the pivotal questions shaping the consumer AI battle, focusing on OpenAI's internal GitHub-like tool, Meta's new applied AI engineering org, Amazon's AI ad ambitions, and the geopolitical chip export caps affecting AI scaling. It highlights OpenAI's GPT‑5.3...
From AI Experimentation to Enterprise Impact at GTC
Heading to @nvidia #GTC. I’m looking beyond the headlines to what actually matters for enterprises: • AI infrastructure for production workloads • Model efficiency + cost economics • Agentic systems in real workflows • Governance, security, and risk • Customer deployments at scale AI is shifting...
What AI in the Back Office Really Means for Life on the Shop Floor
Retail back‑office AI is reshaping security by turning passive camera feeds into proactive threat detectors. Smart cameras now analyze video, audio and behavioral cues to spot loitering, aggression or concealed weapons, while instant alerts route incidents to on‑floor staff. Integrated...
What This Texas Republican Primary Revealed About the Politics of AI Data Centers
Texas Republican primary saw incumbent Agriculture Commissioner Sid Miller lose to businessman Nate Sheets. Miller warned that AI data centers were devouring farmland and water, proposing “agriculture freedom zones.” Voters rejected his message, indicating AI economic momentum may trump rural...
Practical AI for Actuarial Modeling
In this episode, Dale Hall talks with Igor Nikitin, CEO of Nice Technologies, about the practical integration of AI into actuarial modeling, especially for pension and longevity risk transfers. Igor outlines concrete AI use cases such as a second‑reviewer for...
Using AI for Engineering Optimization
MIT researchers have introduced a generative AI system that uses a pre‑trained tabular foundation model as a surrogate within Bayesian optimization. The approach automatically identifies the most influential variables and solves high‑dimensional engineering problems up to 10‑100 times faster than...
Harvey to Integrate with Microsoft 365 Copilot
Harvey announced a deep integration with Microsoft 365 Copilot, embedding its legal intelligence directly into the Copilot environment. The new feature lets lawyers invoke the Harvey Assistant from within Copilot to analyze contracts, research market terms, and pull precedent without...
He Vibe-Coded a Crisis for Higher Education
A 22‑year‑old engineer, Advait Paliwal, released an agentic‑AI tool called Einstein that can retrieve assignments from Canvas, solve complex computer‑science problems, and generate completed PDFs. The service, priced between $40 and $200 per month, attracted over 100,000 site visits before...
CoreWeave Announces Agreement to Power Perplexity’s AI Inference Workloads
CoreWeave and Perplexity have signed a multi‑year strategic partnership to run Perplexity’s inference workloads on CoreWeave Cloud, using dedicated NVIDIA GB200 NVL72 clusters. The deal includes deployment of Perplexity Enterprise Max, which adds advanced search, research, and data‑visualization capabilities for...
AI-Enabled Lesion Quantification Tool Earns New CPT Code
The American Medical Association has granted a Category 3 CPT code (X567T) to AIQ Solutions’ TRAQinform IQ, an AI‑enabled lesion quantification platform cleared by the FDA in 2018. The tool analyzes changes in lesion size and metabolic activity, providing clinicians with more...
Comfy Cloud Is Out of Beta — and It's Just Getting Started
Comfy Cloud announced it has left private beta, delivering the full ComfyUI experience as a hosted service. The platform now includes the majority of custom nodes used in local workflows, on‑demand GPU billing that charges only for active processing, and...
CMiC Launches AI-Powered Construction ERP to Improve Processes
CMiC unveiled NEXUS, an AI‑powered construction ERP that integrates more than 25 intelligent agents, natural language processing, and embedded analytics across its financial and project management suite. The platform automates document handling, drawing classification, submittal creation, and financial reporting, turning...
Grammarly Offering Manuscript Reviews by AI Versions of Recently Deceased Professors
Grammarly has launched an "Expert Review" AI feature that lets users select real scholars—including deceased professors—to critique manuscripts. The tool draws on scraped publications to generate feedback and even rewrites text in the chosen expert's style. Academics quickly condemned the...
Why Data Center Firms Are Working With Trump on Energy Costs
The White House, led by former President Donald Trump, convened executives from Google, Meta, Microsoft, OpenAI and other AI‑focused firms to address rising electricity costs tied to new data centers. The companies signed a voluntary "ratepayer protection pledge" committing to...
Iran’s Authorities Using NtechLab’s Live Facial Recognition to Crush Dissent
Iranian authorities have integrated Russia’s NtechLab FindFace facial‑recognition system into a live surveillance network, accessed through local partners Rasad, Kama and BPO. The technology is deployed in Tehran’s university entrances, Mashhad’s subway, and other public venues to identify and detain...
Microsoft Expands Sovereign Cloud Capabilities to Support AI Models
Microsoft announced a major expansion of its sovereign‑cloud portfolio, adding Foundry Local for on‑device AI inference and making Azure Local and Microsoft 365 Local globally available in disconnected environments. The new services let qualified customers run large AI models and core...
Authentic Brands to Use Seel’s Agentic AI for Post-Purchase Processes
Authentic Brands will integrate Seel’s agentic AI‑driven Worry‑Free Purchase platform to automate post‑purchase issues such as missing packages, claims and refunds. The solution will be deployed across its portfolio, which includes ten brands in Digital Commerce 360’s Top 1000 like Reebok,...
South Korea’s Music Industry Unites to Declare ‘War’ on AI Copyright Infringement
South Korea’s six leading music‑industry bodies have formed a coalition to combat AI‑driven copyright infringement, declaring a state of emergency and a "war" on unauthorized use. The group, chaired by KOMCA president Lee Si‑ha, issued a joint declaration demanding consent‑based...
Podcast Interviews Shaping Startups and Careers
lol heard a 2nd startup today that has made sales and evals based on this podcast its fun to be "just an interviewer", but i'm always mindful/humbled by the fact that we are playing with live ammo here. LS has directly...
Anthropic CEO's Rant Reveals OpenAI's Stronger Position
Imagine thinking that saying FU to the President is going to work out well for you. Seeing so many delusional takes like "Claude's traffic is up, OpenAI is screwed." OpenAI is in a WAY better spot than Anthropic now. Anthropic DEPENDS on...
New AI-Enabled Heart Failure Implant Shows Early Potential
Relief Cardiovascular unveiled its AI‑enabled Relief System, an implantable valve and sensor that modulates inferior vena cava pressure to boost renal perfusion in diuretic‑resistant heart‑failure patients. In the first‑in‑human RELIEF‑FIH study of eight European participants, the device achieved 100% procedural...
Nano Banana 2 Shines Brightest Within Lovart AI
Nano Banana 2 is powerful, but using it inside @lovart_ai is the real flex 🔥 → lightning fast generation → stunning 4K outputs → post-gen control is wild Let's dive in 🧵 ↓ https://t.co/bPHUKLBcjR
AI Poised to Upend Software, Jeopardizing Private Equity
The Extinction of Private Equity? My latest memo on the potential of AI to disrupt the software industry and its effect on private equity: https://t.co/GMi44JG2PQ https://t.co/oZYyMnfppx
AI and Technology Alone Won't Fix Revenue Cycle Challenges: The Automation Paradox in RCM
Healthcare finance leaders are pouring capital into AI and automation to solve revenue cycle management (RCM) challenges, yet many are discovering that automating flawed workflows only accelerates errors. The emerging "automation paradox" shows that without clean data and standardized processes,...
CIOs Must Evolve: AI Drives Strategic Leadership
AI is rewriting the rulebook for leadership. 🤖📈 @Gartner_inc reveals how CIOs must evolve as AI transforms the workplace. It’s time to move beyond the hype and use AI for real strategic planning and continuous improvement. Read more here: https://t.co/XDAHd3CquP #GartnerIT #CIO #AI...
AI Solves Don Knuth's Long‑Standing Open Problem
Math legend Don Knuth is losing his mind because Claude Opus 4.6 just solved an open problem he had been working on for weeks 🤯 https://t.co/elAYEgOQwM
Determining the Roles Required for Effective AI Governance in Higher Ed
Universities are racing to formalize AI governance as generative tools permeate admissions, instruction, research, and operations. Institutions such as Austin Community College and Cornell have created interdisciplinary committees and senior roles—like a vice provost for AI strategy—to define policies, vet...
Ayar Labs to Power 2028 AI with Silicon Photonics
How Ayar Labs is going to ramp silicon photonics into the 2028 generation of AI systems. $NVDA $AMD https://t.co/5PdPsFHxqX
NDSS 2025 – On The Realism Of LiDAR Spoofing Attacks Against Autonomous Driving Vehicle
The NDSS 2025 paper presents the first large‑scale measurement of physical‑world adversarial attacks on commercial traffic‑sign recognition (TSR) systems used in autonomous vehicles. While academic attack techniques can achieve 100% success against specific commercial TSR functions, overall success rates are...