Today's AI Pulse
OpenAI unveils ChatGPT Images 2.0 with multilingual text and ‘Thinking’ mode
OpenAI rolled out ChatGPT Images 2.0, a next‑generation image model that adds multilingual typography, real‑time web research, and agentic reasoning. The new ‘Thinking’ mode lets paid users generate up to eight coherent images per prompt, including floor plans, maps and UI mock‑ups. The feature expands the model’s ability to work with uploaded content.
Also developing:
AI Literacy Workshop: Evidence-Based Session From a Futurist — Get AI-Ready in One Day
Futurist Ian Khan warns that most AI literacy workshops are either too theoretical or overly tactical, delivering little lasting impact. He proposes a data‑driven, decision‑focused format that ties AI concepts directly to strategic choices executives must make in the next 90 days. Industry research shows AI‑ready workforces capture three‑to‑five times more value, while interactive sessions boost knowledge retention to 75%. Khan’s workshops include a pre‑diagnostic, proprietary AIRS™ and FutureSHIFT™ frameworks, and post‑session activation to ensure measurable outcomes.
Claude Opus 4.7 Misclassifies Simple Tasks, Lacks Manual Override
I think the adaptive thinking requirement in Claude Opus 4.7 is bad in the ways that all AI effort routers are bad, but magnified by the fact that there is no manual override like in ChatGPT. It regularly decides that non-math/code...
Another Win for Opportunistic Screening: AI Turns Head CT Scans Into Heart Assessments
Researchers at Harvard Medical School have trained deep‑learning models on nearly 100,000 head CT scans to predict cardiovascular disease (CVD) risk and estimate coronary artery calcium (CAC) scores. The AI‑derived CVD timing model outperformed the American Heart Association’s PREVENT risk...
AMD and French Government Announce Plans to Advance AI Innovation, Research and Open Ecosystem Development
AMD and the French government signed a multi‑year Letter of Intent in Paris to deepen collaboration under France’s National AI Strategy. AMD will provide CPUs, GPUs, software and training through its university and developer programs, while partnering with GENCI, the...
Tesla FSD Misreads Exit, Recovers on Tight Ramp
I asked Tesla Self-Driving 14.3.1 to exit the Mason / O'Farrell Garage in Downtown San Francisco. It tries to go up instead of down, so I had to intervene to help it find the right path. But it handled the tight...
Month-by-Month AI Data Scientist Roadmap for 2026
The 2026 AI Data Scientist Career Roadmap: • Month 1: Python, Pandas, Scikit-Learn • Month 2: LangChain + OpenAI • Month 3: LangGraph • Month 4: AWS Cloud  Want help? 👉 Register here: https://learn.business-science.io/registration-ai-workshop-2
AI Documentation Seen as a ‘Competitive Advantage’ in Hospice Regulatory Labyrinth
Hospices are turning to artificial‑intelligence documentation platforms to ease staffing shortages and meet mounting Medicare audit demands. Dr. Brian Haas, national medical director at Ascend Hospice, argues that precise, rapid documentation will become a measurable competitive advantage. While AI can...
Google’s Gemini AI Cuts Scammers, Blocks 8.3B Ads
Google's 2025 Ads Safety Report: With Gemini's help, Google blocked a record 8.3B ads, up 63% YoY, but suspended 36% fewer advertisers, attributing the disparity to its use of AI "Google said the rise in blocked ads also reflects the...
Albertsons Is the New Walmart when It Comes to Tech Dominance
Albertsons is accelerating its tech agenda, deploying AI across merchandising, pricing, inventory and supply‑chain functions. The retailer reported that AI‑driven personalization is boosting conversion rates, basket sizes and loyalty, while generative‑AI scheduling and computer‑vision tools are sharpening demand forecasts and...
Plan Faster with Claw/Hermes, Boost GStack Efficiency
Pro tip for the GFamily - if you use GStack with Claude Code, but also have a Claw/Hermes with GBrain... I like to do my GStack planning (autoplan skill) in Claw/Hermes since it's faster, and then drop the plan and...
AI Shifts IT Roles From Operator to Orchestrator
A new SolarWinds report shows 80% of IT professionals have moved from hands‑on operations to overseeing AI‑driven automation, redefining their role as orchestrators. Respondents cite reduced manual effort (65%), faster root‑cause analysis (61%) and higher decision confidence (49%) as tangible...
Claude Usage Rose by over 40% Amid Increased Attention Last Month, but Remains Far Behind ChatGPT
Claude, Anthropic’s AI chatbot, saw US usage jump over 40% in March, climbing from 2.96% to a peak of 4.38% before settling at 4.27% in early April. The surge translates to several million new weekly users, according to the poll’s...
Salesforce, Databricks Unveil Agentic AI Governance Tools
Salesforce and Databricks introduced new agentic AI governance features, extending their Agent Fabric and AI Gateway platforms. The tools provide granular token management, permission controls, fixed handoff rules, and API usage policies to curb the rapid spread of AI agents...
AI + Math Learning. How to Solve a New Problem
The National Council of Teachers of Mathematics (NCTM) is urging teachers, principals and district leaders to stay current on AI trends as artificial‑intelligence tools become commonplace in math classrooms. While AI can deliver instant step‑by‑step solutions, the organization warns that...
True Positive Weekly #157
The latest True Positive Weekly roundup spotlights Stanford’s 2026 AI Index Report, a BBC piece on a monkey selfie as a guard against AI‑generated misinformation, and Google’s Gemini now pulling personal photos to craft bespoke AI images. It also highlights...
TBM 417: Before You Fire All Your Glue People Because of AI
The article warns against indiscriminately firing "glue people"—those who handle coordination, judgment, and informal knowledge—by assuming AI can replace their work. It explains that AI excels at automating repetitive, friction‑laden tasks but fails to capture the nuanced decision‑making, trust‑building, and...
Hugging Face Pushes Into “Computer Use” With HoloTab Agent that Works Through Your Browser
Hugging Face unveiled HoloTab, a Chrome extension that embeds its new Holo3-35B-A3B model to perform “computer use” tasks directly in the browser. The agent can click, type and navigate web pages, handling multi-step workflows without any site-specific integration. HoloTab’s performance...
OpenAI Debuts GPT-Rosalind, a New Limited Access Model for Life Sciences, and Broader Codex Plugin on Github
OpenAI unveiled GPT‑Rosalind, a domain‑specific reasoning model built to accelerate life‑science research, alongside a Codex plugin that links the model to over 50 public multi‑omics databases. The model demonstrated top‑tier performance on benchmarks such as BixBench and LABBench2, surpassing GPT‑5.4...
Grokstream Expands Scope and Reach of AIOps Platform
Grokstream has upgraded its Grok AIOps platform with a Proactive Problem Identification engine that uses predictive and causal AI to cluster recurring anomalies, pinpoint root causes and recommend automation. The release also introduces GrokGuru, a generative AI assistant that captures...
Inventors Have Looked to Perfect ‘Robot Umpires’ for Better Part of a Century
Major League Baseball officially launched an automated ball‑strike (ABS) system this season, ending a century‑long quest for robot umpires that began with a 1938 patent. The technology gives each team two challenges per game, letting players contest calls by tapping...
Computer Says Kill: Collapsing the Chain W/ with Matt Mahmoudi
In this inaugural episode of the "Computer Says Kill" series, Alex Dunn talks with Matt Mahmoudi, an assistant professor of digital humanities at Cambridge and adviser to Amnesty International, about the emerging use of large language models (LLMs) in military...
Google, Pentagon Discuss Classified AI Deal
Google is in talks with the U.S. Department of Defense to let the Pentagon use its Gemini AI models in classified environments. The discussions include a contract that would permit all lawful defense applications while adding clauses to block domestic...
MediaPipe Gives 3D Single-Person, YOLOv26 Multi-Person 2D
MediaPipe Pose vs YOLOv26 Pose — two differences that change everything: → Single person vs multi-person → Relative 3D vs 2D only MediaPipe: locks on one person, gives 3D landmarks, runs on phones. YOLOv26: detects everyone, but 2D keypoints only. Same task. Different philosophy. #ComputerVision #PoseEstimation...
AI Is Bringing Added Complexity for HPC Sites. How Are They Handling It?
AI workloads have transformed CINECA’s supercomputing environment, adding layers of software, storage and orchestration beyond traditional HPC. Since 2023 the Italian consortium expanded from a single Slurm‑based stack on Lustre to include Kubernetes, OpenStack, S3 object stores and AI‑focused frameworks...
swXtch.io Introduces the swXtch AI Router for Live Media Workflows
swXtch.io launched a new brand, swXtch.ai, and unveiled the swXtch AI Router, a chat‑driven platform that turns operator intent into production‑ready AI pipelines for live audio and video. The Router integrates with Microsoft Fabric and NVIDIA AI infrastructure, allowing any...
The only Way to Fight Deepfakes Is by Making Deepfakes
Deep‑fake detection firms such as Reality Defender, Pindrop and GetReal are racing to combat AI‑generated audio, video and image fraud, a market now valued at roughly $5.5 billion. These companies train detection models by creating their own deepfakes, using a student‑teacher...
AIDURIX: The Living Navigation System That Helps Hotels Actually Move Forward With AI
AIDURIX launches a living navigation system designed to bridge the gap between AI ambition and practical adoption in hotels. The platform tailors AI roadmaps to each property's size, culture, and leadership capacity, emphasizing fair process and continuous reinforcement. By positioning...
Fibr AI Launches Ads-to-Web and LLM-to-Web Personalization to Convert Paid and AI-Referred Traffic
Fibr AI unveiled Ads‑to‑Web and LLM‑to‑Web personalization, technologies that carry the intent expressed in paid ads or AI‑generated queries through to the landing page. The platform can spin up to 1,500 tailored page variants in under two weeks, delivering 35‑50%...
Testlio Takes On AI Chatbot Risk Before It Reaches Customers
Testlio unveiled its AI Chatbot Testing solution, a human‑led service that evaluates chatbots across four risk domains—safety, consistency, accuracy, and user experience. The offering uses the proprietary LeoPulse score to benchmark release readiness, weighting safety, reliability, and capability. Early adopters,...
An AI View of SentinelOne
McKinsey warns that AI is enabling cybercriminals to launch attacks in minutes rather than days. Attackers leverage AI for hyper‑personalized phishing, deepfakes, instant malicious code, and adaptive tactics. In response, 77% of organizations have integrated AI into security functions, automating...
Agentic AI Security Needs Layered, Integrated Defenses
We keep asking how to solve agentic security as if there is a single answer, but most of the conversations I am having suggest it is a combination of least privilege, access controls, monitoring, and good architecture. The question might not...
Britni Ihle Joins BlackRock to Lead Trading Transformation in AI Push
BlackRock has hired Britni Ihle as Managing Director to lead its Trading Transformation within the Global Markets AI team. Ihle arrives from Citi, where she oversaw cross‑product and credit quant trading, and previously held senior systematic trading roles at Credit...
From Legacy Processes to AI-Native Work
The article argues that AI adoption in knowledge work is hindered more by organizational design than by technology itself. Companies must replace legacy processes with AI‑native orchestration models that blend human roles and intelligent agents. A key obstacle is the...
CEO-Board Survey Finds Big Workforce Changes Ahead Due To AI
A new survey of 1,200 CEOs and board directors finds artificial intelligence will trigger the most significant workforce transformation in a decade. Over two‑thirds say AI will fundamentally alter job functions within three years, and 45% anticipate headcount cuts as...
New GStack Gate Stops AI Guesswork at Critical Forks
Karpathy's Confusion Protocol is now in GStack Karpathy called it: the #1 AI coding failure mode is the agent confidently picking the wrong path at an ambiguous decision point. You lose 10 minutes of work and have to start over. gstack now...
Walmart Backs AI-Driven 3D Weaving for US Manufacturing
Brands such as Walmart signed letters of support for unspun’s plan to build domestic manufacturing capacity in the United States using an AI-enabled 3D weaving technology. https://t.co/Dv03xpHLQE #sustainability #AI #5G #cloud #edge #futureofwork #digitaltransformation https://t.co/1VS4PYadIK
Employer Branding & Recruitment: The Role of AI Video Content in Hiring
AI‑generated video models such as Sora 2 and Kling are reshaping employer branding, but their outputs often contain watermarks, noise, or color inconsistencies that can undermine professionalism. A suite of specialized tools—including Video Watermark Remover AI, Topaz Video AI, DaVinci Resolve,...
Spot the AI Formula: Copy, Contrarian, Obvious Explanation
Is there a clearer tell for an LLM-generated post than: [What everyone's doing] [Contrarian comment] "Here's why:" [Obvious explanation]
OpenAI Unveils GPT‑Rosalind for Pharma Drug Discovery
I truly believe OpenAI is hoping to drive a ton of revenue based on what it can do with health/medical -> OpenAI launches GPT-Rosalind, an AI model for life sciences research, including drug discovery, as a research preview for customers...
Stellantis and Microsoft Expand AI Collaboration Across Operations
Stellantis and Microsoft have launched a five‑year partnership to co‑develop more than 100 AI initiatives covering customer care, product development and operations, alongside AI‑driven cybersecurity, Azure cloud modernization, and a company‑wide rollout of Copilot tools. The automaker aims to shrink...
AI Generates Real‑Time Images From Your Sketches
WHAAAT?! AI can now generate photos of whatever you're drawing IN REAL TIME 🤯 https://t.co/RwK3wVhHKK
Loved PyCon DE: AI Community, Now on Family Break
Had a great time at PyCon & PyData DE. Highly recommend it. Great open-source, community-focused conference with lots of builders in the Python AI, LLM and agent space. Taking a short family break, my first "vacation" in years (hopefully, I won't...
OpenAI Debuts GPT‑Rosalind, Entering Biopharma After Anthropic
NEW: OpenAI is the latest tech giant to move into biopharma, launching Thursday GPT-Rosalind, a life sciences-tailored version of its LLM Trails behind the very similar launch of Anthropic's Claude for Life Sciences by ~5 months. More here: https://t.co/zkXDwY0BBI
From Internal Tools to an Agent Factory
I started by building agents because of internal product work and OpenClaw. Now, it feels like I’m building an agent factory.
AI Accelerates Mid-Market Data Integration for Faster Decisions
Contributor Spotlight: Henry Park (p. 33): AI makes mid-market data integration faster and more accessible - connect systems, improve insight, speed decisions. https://t.co/YrxFqMpTXp #AI #Data #SIOP https://t.co/l3pRT7Y3Zz
New Multi‑Region Endpoint Boosts Availability, Preserves Data Residency
You want @AnthropicAI Claude Opus 4.7 served from a single @googlecloud region? Got it. Oh, you want the global endpoint for max availability? Do that. What's new is the multi-region endpoint with more availability, but respecting data residency. https://t.co/CSUJhvOCMX
Venice Uncensored 1.2 Tuned on Targon Subnet 4
We've started working with some Bittensor subnets. New Venice Uncensored 1.2 model was tuned on Targon/subnet4
AI's Transparency Gap Hinders Theory Verification
The problem with AI right now is that we do not have enough transparency to prove or disprove this theory. I just wrote about this. I don’t know what the real issue is or hopefully was.
From One AI-Generated Page to Cohesive Websites
You can get AI to write a page in seconds. The hard part is making the next five pages feel like they belong. Tomorrow at 10am PT I’ll show the exact workflow I use to go from one strong page...
New AI TUIs May Expand Attack Surface via React
I know there are some new TUIs coming out for AI development but I’m concerned about increasing the attack surface with web technologies like react. Securing AI is hard enough as it is.