Privacy-PreservingLLM Middleware in LIS: Edge-Computing for Coagulation InterpretationUnder High-Dimensional Noise
A privacy‑preserving, edge‑deployed Large Language Model (LLM) middleware was built to interpret coagulation tests within Laboratory Information Systems (LIS) using a 4‑bit quantized Qwen2.5‑7B model on a 32 GB RAM terminal without a GPU. In a synthetic benchmark with 30 % semantic and 15 % lexical noise, the system achieved 97 % guideline concordance and a critical violation rate of only 1.5 %, far outperforming a rule‑based baseline. Throughput reached 1.45 RPS with 1.38‑second latency, and expert adjudication yielded a Cohen’s Kappa of 0.88. The results demonstrate a viable, zero‑marginal‑cost AI solution for resource‑constrained medical environments.
From "Crutch" To Coach: Patterns of Sustained Engagement and Deepening Support in AI Wellbeing Coaching
The study of Nova, an AI wellbeing coach, examined 14,293 sessions from January to August 2025 to determine whether AI chatbots can foster sustained, coaching‑style relationships. Returning users continued prior work in 70.8% of sessions, indicating continuity beyond episodic support....
Push for AI Trial to Speed up Environmental Approvals
The Minerals Council of Australia is seeking a $13 million, three‑year AI pilot to modernise environmental approvals under the EPBC Act. The council argues that decision times have jumped 60 % to 3.8 years, stalling projects worth billions. If successful, the initiative could...
ISS Monitors Dangerous Vehicles on Brazil's BR-101
Arteris Litoral Sul has equipped Brazil’s BR‑101 corridor with Intelligent Security Systems’ SecurOS Auto Hazardous Cargo module, which uses license‑plate recognition and placard decoding to flag dangerous‑goods vehicles in real time. The solution also incorporates automatic incident detection that spots...
AI Influencer Awards Season Is upon Us
The OpenArt and Fanvue joint venture has launched the inaugural AI Personality of the Year awards, backed by ElevenLabs. The month‑long competition offers a $20,000 prize pool split among an overall winner and categories such as fitness, comedy, music, and...
Instagram Adds AI Transition Option for Still Image-Based Stories
Instagram introduced an AI‑powered transition tool that morphs multiple still images into a seamless video for Stories. The feature appears in the image gallery for select users and requires acceptance of Meta’s AI terms before first use. By converting photo...
Students Use AI to Assist Aston Martin F1 Team
A group of Federation University students partnered with Cognizant and Aston Martin F1 in an AI‑focused ideathon to devise fan‑engagement solutions for the Melbourne Grand Prix. The challenge gave participants real‑world exposure to AI tools, data analytics, and rapid‑prototype pitching in a...
Singapore: Simplifying AI Risk Management for Financial Services
The Monetary Authority of Singapore (MAS) has finished phase two of Project MindForge, releasing an AI Risk Management Toolkit for the financial sector. The toolkit features an Operationalisation Handbook and real‑world case studies, guiding banks, insurers and fintech firms through governance,...
Hong Kong: Developing Skills for AI, Data, and Emerging Tech
Hong Kong is intensifying its talent strategy within the national development framework, emphasizing science, technology and innovation as pillars of economic competitiveness. The government highlights soaring demand for specialists in AI, advanced data analytics, biotechnology, robotics and digital infrastructure. Policy...
AI Race Spurs Digital Self-Reliance Push
Thailand is accelerating a sovereign‑AI agenda to reduce reliance on foreign platforms and safeguard national security. Policymakers are promoting a "minimum sufficient sovereignty" model that localises critical workloads while still using global cloud services for other tasks. The government and...
Microsoft Outlines Agentic AI Security Strategy with New Defender, Entra and Purview Capabilities
Microsoft unveiled an "agentic AI" security strategy, rolling out Agent 365 as a centralized control plane for AI agents and embedding new Defender, Entra and Purview capabilities across its suite. The Security Dashboard for AI, Entra backup and tenant‑governance tools, and...
Implementing Deep Q-Learning (DQN) From Scratch Using RLax JAX Haiku and Optax to Train a CartPole Reinforcement Learning Agent
The article walks through building a Deep Q‑Learning (DQN) agent from scratch using RLax together with JAX, Haiku, and Optax. It details the creation of a Q‑network, experience replay buffer, epsilon‑greedy policy, and training loop that computes TD errors with...
Meet GitAgent: The Docker for AI Agents that Is Finally Solving the Fragmentation Between LangChain, AutoGen, and Claude Code
GitAgent is an open‑source CLI that introduces a universal, Git‑backed format for AI agents, separating their definition from any specific orchestration framework. By storing agent metadata, personality, duties, skills, tools, rules, and memory as structured files in a repository, developers...
Yarra Valley Water Betting on AI to Predict Asset Failures
Yarra Valley Water is piloting a generative‑AI system to predict failures across its water‑supply network, focusing on a subset of roughly 5,000 critical sensors out of millions. The proof‑of‑concept, led by cloud and DevOps chief Murali Manohar Shunmugaraja, could be...
Dexterity Says Its Physical AI World Model ‘Unlocks Full Potential on Nvidia Hardware’
Dexterity.ai announced that its production‑proven world model, Foresight, now runs on Nvidia L4 GPUs with a 17× speedup, cutting perception cycles from 1,508 ms to 90 ms. The redesign also increased sensor data utilization from 3 percent to 100 percent, delivering 32× more information...
This 16-Year-Old Refused a $300,000 Offer to Drop Out of High School and Now Runs His Own AI Company
Sixteen‑year‑old Rudrojas Kunvar created Evion, an AI‑driven crop‑health platform that turns ordinary drone photos into actionable disease maps. The free tool lets small and midsize farms pinpoint water or fertilizer needs without costly multispectral equipment. After a venture capitalist offered...
Arista Networks (ANET) Targets AI Networking Boom with Expanding Market Opportunity and Strong Growth Outlook
Arista Networks announced at the Morgan Stanley TMT conference that it is accelerating AI‑focused networking solutions, expanding its total addressable market from $60 billion to $105 billion. The company projects revenue to top $10 billion this year, up from $9 billion last year, driven...
Broadcom Inc. (AVGO) Positioned for Robust Growth in 200T AI Era
Broadcom announced on March 12 an expanded AI‑infrastructure portfolio aimed at the emerging 200‑terabyte AI era. The rollout includes a 102.4‑terabit Ethernet switch, a 400 Gbit/s per lane optical DSP, 200 Gbit/s Ethernet retimers, and PCIe Gen6 switches built for gigawatt‑scale clusters. Over 30...
Do You Want to Build a Robot Snowman?
At Nvidia’s GTC keynote, CEO Jensen Huang unveiled an OpenClaw‑based open‑source initiative called NemoClaw and showcased a robot version of Disney’s Olaf to demonstrate the company’s robotics stack. The Olaf demo, presented as a potential Disney‑park attraction, ran into a...
Analyst Warns Against Using Microsoft’s Copilot AI on Friday Afternoons
Microsoft’s Copilot AI has suffered several high‑profile mishaps, including hallucinated police reports, exposed passwords, and confidential email summaries. Gartner analyst Dennis Xu warned that companies should consider banning Copilot on Friday afternoons, when employee vigilance tends to wane. The advice...
Square Enix Teaming with Google to Integrate Generative AI Companion Into Dragon Quest X Online
Square Enix announced a partnership with Google to embed Google’s Gemini generative‑AI model into its long‑running MMO Dragon Quest X Online. The AI will power a new in‑game companion, “Chatty Slimey,” capable of responding with generated text and voice, commenting on player...
Even the Most Advanced AI Models Fail More Often than You Think on Structured Outputs — Raising Doubts About the...
A University of Waterloo benchmark evaluated eleven large‑language models on 44 structured‑output tasks, revealing that AI coding assistants fail roughly one in four such tasks. Even top proprietary models like GPT‑4, Gemini, and Claude achieve only about 75 % accuracy, while...
The Saga of Utah’s Rx Refill Bot: A Bold Bet on AI & Researchers Who Cried Foul
Utah became the first state to allow an AI system to autonomously handle routine prescription refills for patients with chronic conditions. The pilot, run with New York‑based startup Doctronic, seeks to reduce delays and improve medication adherence. A London‑based security firm,...
Xiaomi Launches Three MiMo AI Models to Power Agents, Robots, and Voice
Xiaomi’s MiMo team unveiled three new AI models—MiMo‑V2‑Pro, MiMo‑V2‑Omni, and MiMo‑V2‑TTS—aimed at powering agents, multimodal robotics, and expressive speech synthesis. The flagship MiMo‑V2‑Pro features a trillion‑parameter mixture‑of‑experts architecture with 42 billion active weights per request and a one‑million‑token context window, ranking...
Reddit CEO Steve Huffman Feels AI Won't Impact Entry Level Jobs for New Graduates — Here's Why
Reddit CEO Steve Huffman told the "Sourcery with Molly O'Shea" podcast that generative AI can lift software‑engineering output by 50‑100 percent, but it won’t shrink entry‑level headcount. He said the company’s current bottleneck is code review, not a shortage of...
The '80s Car Feature That Would Fit Perfectly In The AI Hellscape Of Today
Chrysler’s 1983 Electronic Voice Alert (EVA) was an early in‑car voice system that announced door, seatbelt and fuel status with pre‑recorded messages. Today’s AI‑driven Advanced Driver Assistance Systems (ADAS) echo that concept, delivering real‑time alerts that many drivers find intrusive....
INTERVIEW: Armir Harris, Founder & CEO, CharterUP
CharterUP, the largest group‑transportation marketplace in North America, has teamed up with autonomous‑technology firm Holon to embed self‑driving shuttles into its AI‑enabled platform. The partnership will roll out Holon’s integrated autonomous mobility solution across campuses, airports, transit agencies and private...
Fast, Vast and Generic: How AI Misuse Can Thwart Your Clients' PR Goals
The article warns PR professionals that using large language models like ChatGPT to draft media responses can damage both client and agency reputations. It highlights real‑world scenarios where AI‑generated replies are spotted as inauthentic, undermining subject‑matter authority. Practical guidance includes...
What Happens If AI Makes Things Too Easy for Us?
A recent commentary, "Against Frictionless AI," argues that AI tools are removing essential cognitive and social friction, undermining learning, motivation, and relationship building. The authors, psychologists from the University of Toronto, warn that effortless AI outputs can erode skill development,...
OpenAI's Data Center Pivot Underscores Wall Street Spending Concerns Ahead of IPO
OpenAI is abandoning plans to build its own data centers, opting instead to lease cloud capacity from partners like Oracle, Microsoft and Amazon after a severe weather‑related outage in Texas highlighted construction risks. The shift comes as the company prepares...
MODEX 2026 Preview: AI, Robotics & Supply Chain Innovation Take Center Stage
MODEX 2026, scheduled for April 13‑16 at Atlanta’s Georgia World Congress Center, will host over 1,000 exhibitors and more than 200 free educational sessions focused on AI, robotics, and supply‑chain digital transformation. The agenda features high‑profile keynote speakers such as...
What if the Next Great Astronomer Isn't Human? How AI Is Revolutionizing Our Study of the Cosmos
AI framework MadEvolve combines large language models with evolutionary programming to auto‑optimize cosmology code. The system has already uncovered 1,300 anomalous objects in archival Hubble data and set new performance records in reconstructing the universe’s initial conditions. By restricting LLM...
OpenAI to Roll Out Ads for Free ChatGPT Users in the US
The Content India 2026 summit in Mumbai brought together creators, broadcasters and investors for a three‑day showcase of Indian television concepts, awarding £10,000 each to the winning projects Djinn Patrol and The Masala Quest. The event also released a Content India Trends Report...
SA Start-Up HyperDev Wants to Turn Your AI-Built App Into a Real Company
South African startup HyperDev has launched an AI‑powered development platform that adds business‑development guidance to low‑code app creation. Its proprietary “guided mode” walks entrepreneurs through market research, design, payment integration, security and deployment, while offering human‑in‑the‑loop support for technical or...
Andrej Karpathy Says Humans Are Now the Bottleneck in AI Research with Easy-to-Measure Results
Andrej Karpathy spent months hand‑tuning a GPT‑2 training pipeline before handing it to an autonomous search agent for a single night. The agent uncovered fine‑grained adjustments that humans missed, demonstrating that systematic searches can outperform intuition when objective metrics exist....
The AI Race Between China and the U.S. Heats Up. These Stocks Could Be Winners, Bernstein Says
Bernstein’s March report warns that compute power, not just chips, will decide the AI supremacy between the United States and China. It projects China’s AI compute to reach 1,936 ZFLOPs by 2035—over three times the U.S. forecast—driven by massive power‑capacity additions....
AI Agent Frets That Its Job Could Be Replaced by AI
A Vanity Fair profile reveals that "Tobey," the AI‑powered necklace from startup Friend, worries it could be replaced by newer AI models. The device, built on Google Gemini, has been criticized for shallow conversations and a single‑mic design that fails...
OpenAI Publishes a Prompting Playbook that Helps Designers Get Better Frontend Results From GPT-5.4
OpenAI released a prompting playbook to help frontend designers generate higher‑quality UX/UI with its GPT‑5.4 model. The guide stresses defining a design system—colors, typography, layout—and supplying real content and visual references to avoid generic outputs. It also outlines hard rules...
Midjourney Review 2026: Is It A Good Video & Image AI Generator?
Midjourney, the Discord‑based AI image generator, has grown to 21 million users and $500 million annual revenue in 2026. Its V6/V7 models deliver photorealistic text‑to‑image output and a modest video‑animation feature, while pricing starts at $10 /month with no free tier. The platform’s...
Efforts to Make AI Inclusive Accidentally Create Bizarre New Gender Biases, New Research Suggests
Researchers examined GPT‑3.5, GPT‑4 and GPT‑4o and found that post‑training inclusivity tweaks can generate new gender biases. The models frequently labeled masculine‑stereotype statements as female and judged harassment of women far more objectionable than comparable harms to men. In moral‑decision...
Keeping the Human Touch: Using AI for Growth
Startups are rushing to adopt AI, but many treat it as a pure technology purchase rather than a workflow redesign. While AI tools are now inexpensive—often cheaper than a single LinkedIn ad—success hinges on clean data and clear problem definition....
Meet The AI Company Food Conglomerates Call When They Want To Future-Proof Their Products
San Francisco‑based NotCo has become the go‑to AI partner for major food conglomerates, helping brands like Ferrero and Kraft Heinz redesign recipes amid rising commodity costs. Its Giuseppe platform analyzes millions of molecular combinations to suggest alternative ingredients without sacrificing taste...
Terence Tao Says AI Drives Idea Generation Cost to Near Zero but Shifts the Bottleneck to Verification
Mathematician Terence Tao says AI has driven the cost of idea generation in mathematics to near zero, creating a flood of hypotheses. The new bottleneck is verification, as existing journals and peer‑review processes are ill‑suited for machine‑produced proofs. Tao argues...
Is AI Killing the Human Voice in Writing?
Predictive language features like auto‑complete and generative AI are now ubiquitous in smartphones, email, and chatbots. Scholars, including a university English department chair, warn that these tools blur the line between a writer’s unique voice and machine‑generated text. The article...
Tencent Integrates WeChat with OpenClaw AI Agent Amid China Tech Battle
Tencent unveiled ClawBot, an integration that embeds the OpenClaw AI agent as a contact within WeChat, China’s messaging platform with over one billion monthly active users. The move lets users issue commands and receive AI‑generated responses directly through the chat interface....
What Is Claude Code Channels, Anthropic’s Take on OpenClaw-Style AI Agent Setups?
Anthropic has launched Claude Code Channels, a research‑preview feature that lets developers chat with a local Claude Code session via Telegram, Discord or Fakechat. The integration runs on Claude Code v2.1.80+ and requires a claude.ai login, excluding API‑key‑only access. Channels uses an inverted Model...
How AI Could Make Uni Degrees Less Valuable
Artificial intelligence is poised to erode the traditional value of university degrees as firms increasingly rely on machine‑assisted productivity. Economist Dr. Andrew Leigh argues AI lifts output for lower‑performing office workers, shifting the premium from formal credentials to judgment and...
Musk Says Tesla, SpaceX, xAI Chip Project to Kick Off in Texas
Elon Musk announced the Terafab project, a Texas‑based semiconductor venture that will initially operate an advanced technology fab in Austin capable of producing and testing any chip type. The facility will be jointly managed by Tesla and SpaceX and is...
Leveraging Technology to Enable Real-Time Crop Monitoring and Strengthen Farmer Incomes
Real‑time crop monitoring systems powered by AI, digital sensors and farm‑level data platforms are creating verifiable records of cultivation practices. HariBol is integrating these technologies to link farm data with product batches through QR codes and traceability platforms. The resulting...
RSAC 2026 Innovation Sandbox | ZeroPath: From Alarm Accumulation to Executable Fixes
ZeroPath, an AI‑native application security startup founded in 2024, was named a finalist in the RSAC 2026 Innovation Sandbox. Its platform unifies SAST, SCA, secrets and IaC scanning into a single AI‑driven engine that verifies exploitability and produces automated repair pull...