Today's AI Pulse
AI agents challenge cybersecurity frameworks, sparking budget hopes and governance concerns
AI agents are being deployed to test existing cybersecurity frameworks, prompting CISOs to see a path to unlock larger cyber budgets. At the same time, industry leaders warn that the lack of clear algorithmic authorization raises governance risks. The convergence of these trends highlights both the promise and the perils of ungoverned AI in security operations.
Also developing:
By the numbers: OneRobotics acquires Nanoleaf for $40M
AI-Powered Stroke Tool Linked to Improved Patient Outcomes in Large Clinical Trial
A large cluster‑randomized trial of more than 21,000 acute ischemic‑stroke patients across 77 Chinese hospitals tested an AI‑powered clinical decision support system (CDSS). The tool, which combines AI‑assisted imaging with guideline‑based treatment prompts, lowered the 3‑month composite vascular‑event rate from 3.9% to 2.9% – a 26% relative reduction – and maintained a 27% reduction at 12 months. Care‑quality scores rose modestly to 91.4% versus 89.8% for usual care, with no increase in mortality or bleeding. Researchers deem the system safe, cost‑effective, and scalable for resource‑constrained settings.
AI-Generated Code Verification Startup Qodo Raises $70M
Qodo, operating as Codium Ltd., announced a $70 million Series B round, bringing its total capital to $120 million. The startup offers an AI‑driven code verification platform that inserts a governance layer into developers' CI/CD pipelines, using specialized agents to detect bugs, compliance...
AI Agents Are Democratizing Finance but Also Redefining Risk
AI agents are now moving capital autonomously, turning complex arbitrage strategies into simple commands and democratizing access to sophisticated finance. Users have seen outsized returns, such as a $300 investment growing to over $2.3 million in four months, thanks to rapid,...
Google's AI Headline Rewrite Test Triggers Publisher Outcry
Google is piloting an AI-driven feature that rewrites the headlines of news articles in Search results, prompting sharp criticism from publishers who say the move oversteps consent and threatens editorial integrity. The dispute highlights growing tension between the search giant...
Savills Teams with Ricoh to Deploy AI‑Driven Drones for Hong Kong Building Inspections
Savills Property Management Group and Ricoh Hong Kong have struck a strategic partnership to roll out AI‑powered drone inspections of building façades across Hong Kong. The solution replaces scaffolding‑based checks, cuts inspection time and delivers data‑driven maintenance insights for the...
Microsoft Adds Anthropic’s Claude to 365 Copilot, Lifts Research Quality 13.8%
Microsoft announced that its 365 Copilot suite now uses Anthropic’s Claude to critique and improve GPT‑generated drafts. The new Critique feature lifts the DRACO benchmark by 13.8%, a move aimed at accelerating Copilot adoption among its 15 million paid enterprise seats.
Fortis Solutions and PG&E Deploy AI Platforms to Modernize Critical Infrastructure
Fortis Solutions introduced its NetRaven and Source of Truth AI-driven platforms, and PG&E began using Atomic Canyon's Neutron generative AI tool at the Diablo Canyon nuclear plant. Both moves aim to cut manual effort, improve compliance, and boost operational agility...
Waymo Co-CEO Pushes Personal-Car Autonomy Amid Austin School-Bus Safety Scrutiny
Waymo co-CEO Dmitri Dolgov announced plans to embed its robotaxi technology in privately owned cars, even as the company grapples with a series of school‑bus‑passing incidents in Austin that have drawn NHTSA and NTSB investigations. The dual narrative highlights both...
Orbital AI's Starcloud Raises $170M, Hits $1.1B Unicorn Valuation for Space‑Based Data Centers
Orbital AI’s Starcloud announced a $170 million financing round that lifts its valuation to $1.1 billion, making it Y Combinator’s quickest path to unicorn status. The funding fuels the launch of its next-generation satellite data center, Starcloud‑2, and underscores growing venture interest...
University of Ottawa Launches New Applied AI Centre for Healthcare
The University of Ottawa has inaugurated the Ottawa Medical Artificial Intelligence Research Institute (OMARI), an applied‑research hub embedded in its Faculty of Medicine and linked to the city’s six hospitals. Led by AI expert Dr. Khaled El Emam, OMARI aims to...
AI Will Supercharge Security, But Transition Will Be Rough
I think the situation with AI and security this talk highlights is kind of indicative of how AI disruption will play out. AI is going to make it really, really easy and cheap to find exploits in software. The end...
Funding Gaps and Talent Shortage Cripple Indian AI Startups
Why Are Indian AI Startups Shutting Down? https://t.co/EsZZ11sDNY #StartupGrowth #StartupLife #startup #startupsoul #artificialintelligence #AI #startupindia #makeinindia
Google’s ‘AI Works for Britain’ Aims to Help Stuck Brits Unlock Progression
Google UK announced the AI Works for Britain programme, a nationwide upskilling effort backed by roughly $2.5 million in Google.org grant funding. The initiative targets the 76% of Britons who feel stuck in their careers by teaching practical AI skills through...
GDF15: AI‑derived Senescence Marker Predicts Mortality
Hopefully blood testing companies can measure GDF15 sooner than later... A deep-learning based biomarker of systemic cellular senescence burden to predict mortality and health outcomes https://t.co/lqM7w3aPfJ
AI-Driven Security Crisis Overwhelms Professionals, Threatens Civilization
Yep, the Computer Security AIpocalypse is happening. Please be patient with the computer security professionals you know for the next few months, as they’ll likely be busy and stressed about preventing the collapse of civilization. https://t.co/AxuD2hEfBD
1inch Launches AI‑agent API Suite, Letting Bots Execute DeFi Swaps Directly
1inch announced that AI agents can now access its full suite of 15 APIs via the 1inch MCP, enabling real‑time swap execution, portfolio analysis and on‑chain transactions. The move positions the DeFi platform as the backbone for the emerging “agent...
Edge AI Cuts Latency by Bringing Compute Closer
From #GTC26: AI performance depends heavily on architecture. @TMobileBusiness and NVIDIA are exploring how edge compute can reduce latency by processing data closer to where it’s created. ⚡ https://t.co/DwZzkJUOUm T-Mobile for Business Partner #EdgeAI https://t.co/ldQpk1r5j2
Even Advanced AI Like DALL‑E Still Miss Basics
amazing. i was giving examples like this for DALL-E three years ago. systems are still struggling with some basics.
Marketing in the Agentic Era
On April 9 at 10:00 AM PDT, former CMO Lena Waters will host a new version of Theory Ventures’ Office Hours. Waters, who led marketing at Notion, Grammarly and DocuSign, will discuss how companies can redesign their marketing organizations for an AI‑driven world. The...
The Hidden Costs of ‘Helpful’ AI
Researchers built a collaborative chess game pairing a strong AI with a weaker, human‑like partner, flipping a coin to decide who moves next. The experiment showed that teams with the compatible, supportive AI consistently outperformed those led by the superhuman...
Yale‑Goldman Alum Builds AI to Automate Pitch Decks
The final boss is Chadwick Bates, Yale grad, Goldman alum, now working on an AI wrapper to automate pitch deck creation
Wiise Boosts A/NZ Partner Program with AI and Product Additions
Wiise has revamped its Australia‑New Zealand partner program, adding AI‑focused training and new prospecting tools. The initiative includes an AI Academy of practical webinars and AI‑powered utilities to help partners qualify opportunities faster. It also introduces partner community days, roadmap visibility,...
California Governor Issues Executive Order About AI
California Governor Gavin Newsom signed an executive order tightening AI procurement standards and requiring new safeguards for companies seeking state contracts. The order expands the state’s existing AI transparency framework, mandating vetting of bias controls, illegal‑content safeguards, and civil‑rights protections....
AI NFL Referees Ignite Election‑season Labor Controversy
I can see A.I.-assisted NFL officiating turning into a big brouhaha right in the midst of election season. I spell out the scenario here… https://puck.news/the-nfls-ai-referee-plan-has-a-labor-problem/
Responsible AI Agents: Prioritize Security, Durability, Availability
When Opus 4.5 came out, it was a one-way door to a new way of engineering. Agents now do most of our coding. Knowing the inherent flaws and over-confidence of LLMs, we sent a clear message to our teams. Vibing and...
Vietnam: AI for Early Detection of Non-Communicable Diseases
Vietnam is fast‑tracking AI integration into its healthcare system to improve early detection of non‑communicable diseases such as cancer, cardiovascular disease, diabetes and chronic respiratory illnesses. The government and hospitals are expanding electronic health records and centralised data platforms, enabling...
Popular AI Gateway Startup LiteLLM Ditches Controversial Startup Delve
LiteLLM, a widely used AI gateway for developers, announced it is ending its partnership with compliance startup Delve and will pursue new certifications through Vanta and an independent auditor. The move follows a credential‑stealing malware breach that exposed weaknesses in...
World Models, Open Claw, and Robots. Discussing AI with Valentino Megale
AI thought leader Valentino Megale outlines three emerging trends reshaping enterprise technology: the open‑source agent orchestration platform OpenClaw, now extended by Nvidia’s security‑focused NemoClaw; the rise of World Models such as LeCun’s JEPA that simulate causal environments rather than merely predicting...
Hong Kong: AI-Driven Innovation and Smart Manufacturing
The Hong Kong Productivity Council (HKPC) is spearheading AI‑driven innovation to boost manufacturing competitiveness and support international expansion, especially through Hong Kong‑mainland China collaboration. Its programme spans frontier research, technology commercialisation, talent development and enterprise transformation, targeting smart manufacturing, robotics,...
5 Reasons Your Claude Skills Keep Breaking (and You Don’t Even Notice)
The post outlines five common pitfalls that cause Claude Cowork skills to underperform, such as never testing Claude without the skill, writing instructions for imagined problems, and patching instead of rebuilding. It introduces a self‑contained skill that automates a baseline‑build‑refine...
Australia: AI Facial Analysis Strengthens Forensic Investigations
Australian researchers at Murdoch University have created a face similarity linkage (FSL) system that quantifies facial geometry to uncover connections between crime victims and suspects. By extracting landmark ratios, the tool normalises variations in angle, lighting and scale, delivering more...
EE TV Rolls Out AI-Powered Smart Search
EE has introduced Smart Search, an AI‑driven content discovery tool for its EE TV platform, allowing users to locate shows and movies by describing mood, theme, plot elements or memorable lines. The feature aggregates results from live TV, on‑demand libraries,...
Malaysia: Navigating an AI-Driven Media Landscape
Malaysia’s media ecosystem is rapidly evolving as smartphones and mobile internet become the dominant gateways to news, pushing traditional broadcast and print toward digital platforms. Social media algorithms now dictate content visibility, while artificial intelligence drives personalization, recommendation and moderation...
Why AI Sandboxes Matter for Responsible Innovation and Public Trust
AI regulatory sandboxes are emerging worldwide as structured testbeds for emerging technologies. Three primary models—regulatory, operational, and hybrid—offer varying degrees of oversight and infrastructure. These sandboxes intervene at different stages of policy development, often using waivers to permit experimentation before...
Microsoft Reduces Copilot Integrations in Windows 11
Microsoft announced a rollback of Copilot features in Windows 11, removing the AI assistant from core apps such as Snipping Tool, Photos, Widgets and Notepad. The move is part of a broader quality overhaul that emphasizes performance, reliability and what the...
A Game Plan for the AI Boom
AlphaGo’s 2016 victory over Go champion Lee Sedol demonstrated that pairing a move‑generation network with a separate evaluation network, trained via reinforcement learning, could master a problem once deemed unsolvable. The same two‑step architecture now underpins today’s reasoning models, which solve...
Microsoft AI Releases Harrier-OSS-V1: A New Family of Multilingual Embedding Models Hitting SOTA on Multilingual MTEB V2
Microsoft released Harrier-OSS-v1, a trio of multilingual embedding models ranging from 270 million to 27 billion parameters. The models use decoder‑only architectures with last‑token pooling and achieve state‑of‑the‑art results on the Multilingual MTEB v2 benchmark. They support a 32,768‑token context window, enabling...
Three-in-One Diode Integrates Sensing, Memory and Processing for Smart Cameras
Researchers at the University of Science and Technology of China have created a single semiconductor diode that simultaneously senses light, stores data and performs processing. By inserting an aluminum‑gallium‑nitride layer into a GaN p‑n junction, the device can switch among...
Last Week in ConTech - 30 March 2026
Construction is entering an AI‑first era where new firms embed technology directly into delivery processes rather than selling standalone software. Recent stealth launch Zero RFI ($13.8 M) and Unlimited Industries ($12 M) illustrate investor appetite for vertically integrated, AI‑native contractors, while private...
One Project, Basic Portfolio Beats PhD Myths
Generative AI Data Scientist Job Myths: • You need months of Math & Stats • You need a PhD • You need Job Experience The reality:  • You need 1 solid AI/ML project • You need a basic portfolio That’s it. Want to learn how? I’m teaching...
Google Gemini’s Lyria Generates Full Songs Automatically
Creating song lyrics with AI is so 2025. Now you can compose entire songs using the Lyria model in Google Gemini. It's pretty darn astonishing. Here's how it works, with sample tracks in a variety of styles: https://t.co/7pNLIXLKWJ #gemini #lyria...
Jerome Powell to Gen Z: Don’t Fear AI—Master It
Federal Reserve Chair Jerome Powell told a crowd of Harvard economics students that the looming AI revolution should be embraced, not feared, by the incoming Gen Z workforce. He highlighted the current low job‑creation environment and warned that large language...
Bipedal Robot Regains Balance After Every Fall
This All-Terrain Bipedal #Robot Keeps Its Balance—and Gets Back Up After Falling via @ZappyZappy7 #AI #MachineLearning #Robotics #ArtificialIntelligence #Innovation #Technology https://t.co/6RlBAo7im4
Continual Learning and Claude Screenshares Accelerate AGI Takeoff
The two remaining inputs for AGI fast take off: 1) continual learning 2) Claude cowork screenshare Millions of Claude users literally teaching the model how to do their job Game. Set. Match
Labeling AI Terms as Slurs Fuels Harmful Anthropomorphism
I actually made a whole video about this whole robot racism stuff and spoke to researchers. Ironically treating words like clanker as a slur contributes to the anthropomorphism of AI, which is a bad thing https://t.co/aoJdfAJOzs
Delaying AI Now Means Falling Decades Behind
Delaying AI by 6 months doesn't put you 6 months behind. It puts you a decade behind. Compounding doesn't pause while you deliberate. By the time most companies feel the gap, it's already irreversible.
Separate AI Execution and Audit Squads Ensure True Accountability
AI agents will tell you the task is done when it isn't. We figured that out fast. Now we run two squads. One executes, one audits. They never talk. The orchestrator sees everything and owns the output. Same accountability structure you'd...
Hugging Face Welcomes GGML Team, Boosting Local AI Runtime
So proud to have @ggerganov and @ggml_org part of the @huggingface team. One of the unsung heroes of AI, powering the most widely used open-source runtime for local AI!
AI Val Kilmer Demands Cooler GPUs, Calls 80°C Inhumane
Val Kilmer’s AI twin is now protesting for 'Better GPU Cooling.' Apparently, 80°C is 'inhumane' for a digital consciousness. I can’t even. 🖥️🔥 #ValKilmer #AIEthics #DigitalTwin #TechSatire #GPU https://t.co/IgSyiy4FRD
Current AI Misalignment Risks Appear Lower Than Predicted
Given that there have been so few cases of harm caused by misaligned AI so far, isn't the solid line more accurately on par with or even ahead of the dashed line? (Acknowledging that the future may change, of course.)