Combining NotebookLM and Obsidian offers a structured approach to managing research, blending AI-driven synthesis with manual curation. Teacher’s Tech explores how NotebookLM’s ability to extract insights from diverse sources, such as PDFs, videos and websites, can complement Obsidian’s focus on long-term knowledge organization.

This presentation highlights recent efforts at the Johns Hopkins Applied Physics Laboratory to advance agentic AI for collaborative robotic teams. It begins by framing the core challenges of enabling autonomy, coordination, and adaptability across heterogeneous systems, then introduces a scalable architecture designed to support agentic behaviors in multi-robot environments.

Skill Leap AI pitted Claude 4.7 Opus against ChatGPT 5.5 across 10 practical scenarios, using Google Gemini as an independent benchmark. ChatGPT stood out in coding, producing clean and consistent output, while Claude held its own in other categories. The head-to-head offers a pragmatic view of which model fits which use case.

Since ChatGPT launched in 2022, top grades have surged in AI-friendly subjects: a UC Berkeley study finds “excellent” marks up 30% in English composition and coding classes, while sculpture and lab courses see no shift. The striking part isn’t A-minus students bumping to A-plus — it’s C-students suddenly landing on A-level.

ArXiv is cracking down on AI-generated junk papers: when there is clear evidence that authors used AI to mass-produce low-quality preprints, the repository will now be able to ban them. The move reflects mounting pressure on the preprint server from low-effort AI submissions. The signal to academia is clear: generative AI as a co-author is fine, AI slop as mass output is not.

Microsoft Research has posted follow-up notes to its paper LLMs Corrupt Your Documents When You Delegate. The researchers clarify what the study actually shows and what it does not: AI agents in delegated workflows do not always stay clean and can quietly alter documents over time.

NASA is testing a next-generation space computer chip that could give spacecraft the ability to operate far more independently in deep space. The radiation-hardened processor is showing performance levels hundreds of times beyond current spaceflight computers while surviving punishing tests designed to mimic the harsh conditions of space.

When generative AI first moved from research labs into business, enterprises accepted a quiet trade-off: capability now, control later. Proprietary data flowed through third-party models with strong results but no real ownership or governance. The article argues that bargain is expiring and companies now need their own data sovereignty, governance, and compliance layer to operate autonomous systems safely.

Exclusive: Doctors say ‘highly concerning’ poll highlights risk to patients of turning to AI for medical advice One in seven people are using AI chatbots for health advice instead of seeing their GP, a UK study has found. The poll of more than 2,000 people found that – of the 15% turning to chatbots – one in four had done so because of long NHS waiting lists.

Most multi-agent systems fail the same way: agents drift apart across handoffs. By turn 3 they are working in different realities. By turn 5 they are repeating each other's mistakes and calling it parallelism.

AWS walks through reinforcement learning with verifiable rewards (RLVR) on SageMaker AI to make reward signals checkable and transparent. The technique works best where outputs can be objectively verified — math reasoning, code generation or symbolic tasks. Layered techniques like Group Relative Policy Optimization (GRPO) and few-shot examples on the GSM8K dataset push accuracy further.

Save to Spotify is a new command-line tool aimed at AI agents like OpenClaw, Claude Code and OpenAI Codex. Users who funnel research through their AI of choice into audio summaries or personal podcasts can route those outputs straight into their Spotify feed. Setup is simple: install the CLI from GitHub, then append "and save to Spotify" to your usual prompt.

Researchers at Yonsei University in Korea have built electronic rings that wirelessly connect to an AI system and translate multiple sign languages into text. Lead researcher Ki Jun Yu calls it a meaningful step toward practical, lightweight, real-world sign-language translation. Earlier camera and computer-vision approaches struggled with lighting changes, fixed setups and interference.

Claude Projects provides a structured way to manage work by creating dedicated AI-powered workspaces that centralize files, instructions and conversations. In his guide, Kevin Stratvert walks through how to get started with this platform, including tips on assigning clear and descriptive project names and organizing tasks into distinct categories like marketing campaigns or research initiatives.

Introducing GridSFM, a small foundation model that can predict AC optimal power flow in milliseconds, boosting efficiency and unlocking cost savings. Learn how GridSFM gives grid operators direct visibility into congestion, stability, and system health. The post GridSFM: A new, small foundation model for the electric grid appeared first on Microsoft Research.

Industry body says energy consumption driven by AI up 15% globally in two years as it warns of societal backlash Datacentres are consuming 6% of electricity in the UK and US, with the growing strain of AI on energy supplies prompting community resistance, according to research. The proportion of electricity used by vast warehouses stacked with microchips to power AI and the internet has risen 15% worldwide in the past two years as annual global investme…

The little pauses, “ums,” and moments when you struggle to find the right word may reveal far more about your brain than anyone realized. Researchers discovered that everyday speech patterns are closely tied to executive function — the mental system that powers memory, planning, focus, and flexible thinking.

Despite years of digitization, organizations capture less than a third of the expected value from digital investments, McKinsey research shows. Most companies start with tech capabilities and bolt apps on top — instead of starting from real customer needs. Customer-back engineering flips that order.

AI now beats traditional weather forecasting in many everyday scenarios — faster, often more accurate, and cheaper to run. But a new study finds that for the cases that matter most — extreme weather, hurricanes, heatwaves — current AI models still fall short. The reason: they are trained on frequent, average patterns and have a blind spot for rare, high-impact events.

Anthropic — the lab whose identity centers on warning about AI risk — says it sees "early signs" of AI not just coding its products but contributing to building itself. Co-founder Jack Clark puts the chance of an AI model fully training its successor by end of 2028 at over 60 percent. The new Anthropic Institute research agenda focuses squarely on this recursive self-improvement loop.

Exclusive: Worker pointed to Iran war and Pentagon’s Anthropic feud as indications the department is ‘not a responsible partner’ Workers developing Google’s artificial intelligence products in the UK have voted to unionize, in part out of concerns about a deal between the company and the US military that was announced last week. In a letter slated to go to management on Tuesday and shared exclusively with the Guardian, workers at Google DeepMind, the co…

Microsoft Research expands MatterSim with faster large-scale simulations and a new multi-task model called MatterSim-MT, which predicts properties beyond potential energy surfaces alone. Conductivity, stability and more come from a single model. A meaningful step in both throughput and scope for AI-driven materials science.

The rise of electricity-guzzling data centers has forced the AI industry to get creative about finding power. Nvidia is teaming up with InfraPartners, Prologis, and nonprofit EPRI to build about 25 micro data centers (5–20 MW each) next to utility substations at five US utilities.

Hermes Agent, developed by Newest Research, is now available as a desktop application, offering a graphical interface that builds on its previous command-line functionality. According to World of AI, the app includes features such as persistent memory, which enables it to retain information across sessions and user modeling, allowing for personalized interactions based on individual […] The post New Hermes Agent Desktop App is Replacing OpenClaw appeare…

World is approaching point where no one can shut down a rogue AI, says director of body behind research It’s the stuff of science fiction cinema, or particularly breathless AI company blogposts: new research finds recent AI systems can independently copy themselves on to other computers. In the doom scenario, this means that when the superintelligent AI goes rogue, it will escape shutdown by seeding itself across the world wide web, lurking outside the…

Creating complex molecules usually requires years of experience and countless decisions, but a new AI system is changing that. Synthegy lets chemists guide synthesis and reaction planning using simple language, while powerful algorithms generate and evaluate possible solutions. The AI doesn’t just compute—it reasons, scoring pathways and explaining which ones make the most sense.

New research from the Federal Reserve Bank of New York confirms what many already suspected: U. spending growth is concentrated almost entirely in the top income tier, fueled by wealth gains from financial assets. Low-income households are squeezed by persistent inflation and have little buffer for additional shocks.

Altman trial, an unusual exhibit drew attention: a trophy inscribed 'Never stop being a jackass. ' OpenAI employees had bought it for researcher Josh Achiam after Musk called him that name. The backstory: Achiam, who worked on AI safety, had questioned Musk's plan to race OpenAI ahead of Google when Musk was leaving the company.

Microsoft researchers share advances in building and operating large-scale distributed systems, spanning datacenters, networking, and the growing intersection with AI during NSDI ’26. The post Microsoft at NSDI 2026: Advances in large-scale networked systems appeared first on Microsoft Research.

Stanford University’s recent research, conducted in collaboration with Tsinghua University, has revealed a surprising shift in how we evaluate the performance of large language models (LLMs). Rather than focusing solely on the architecture of these models, the study emphasizes the importance of the orchestration layer, or “harness,” which coordinates how the model interacts with external […] The post Why Stanford Researchers Say AI Architecture Isn’t th…

Sophomore developer Venkatram is building a local-first alternative to proprietary AI research assistants — essentially NotebookLM running on your own local AI model. The tool aims to turn documents into reusable, searchable assets while preserving the full information content of the original sources. The project is still very early and Venkatram is actively looking for collaborators.

People describe awkward and unnatural process as survey finds nearly half of job seekers have been interviewed by AI Nearly half (47%) of UK job seekers have had an AI interview, research from the hiring platform Greenhouse has found. In its survey of 2,950 active job seekers, including 1,132 UK-based workers, with additional respondents from the US, Germany, Australia and Ireland, it found that 30% of UK candidates had walked away from a hiring process…

For decades, psychologists have debated whether the human mind can be explained by one unified theory or must be broken into separate parts like memory and attention. A recent AI model called Centaur seemed to offer a breakthrough, claiming it could mimic human thinking across 160 different cognitive tasks.

Six facts, no hype, all from the past 60 days. AI is the fastest-growing product category in history. One latest model is so powerful its maker won't release it.

Safe agents don’t guarantee a safe ecosystem of interconnected agents. Microsoft Research examines what breaks when AI agents interact and why network-level risks require new approaches. The post Red-teaming a network of agents: Understanding what breaks when AI agents interact at scale appeared first on Microsoft Research.

In this post, we show how to combine DVC (Data Version Control), Amazon SageMaker AI, and Amazon SageMaker AI MLflow Apps to build end-to-end ML model lineage. We walk through two deployable patterns — dataset-level lineage and record-level lineage — that you can run in your own AWS account using the companion notebooks.

Agensi is a curated marketplace for SKILL. md skills — the folder-plus-instructions format Anthropic created for teaching AI coding agents like Claude Code, Cursor, and Codex new capabilities. Creators publish skills, users install them into their agents.

A Harvard study found AI systems outperformed human doctors in high-pressure emergency medicine triage, diagnosing more accurately in life-or-death moments when patients are first rushed to hospital. Researchers describe the results as a profound shift that could reshape how emergency medicine is practiced.

The release of ChatGPT 5.5 represents a notable step forward in OpenAI’s development of AI systems, addressing key challenges like efficiency and intent preservation. According to Matt Maher, ChatGPT 5.5 achieves a 97.5% accuracy rate in maintaining user intent, matching the benchmark set by Opus 4.7. This improvement, alongside reduced token usage and faster processing […] The post How ChatGPT 5.5 Finally Caught Up to Opus 4.7 in Intent Accuracy appear…

GitHub patched a critical remote code execution vulnerability in under six hours last month. Wiz Research used AI models to surface the bug in GitHub's internal git infrastructure — exploitation would have exposed millions of public and private repositories. The security team reproduced the issue within 40 minutes and shipped a fix the same day.

Researchers have shown that blending quantum computing with AI can dramatically improve predictions of complex, chaotic systems. By letting a quantum computer identify hidden patterns in data, the AI becomes more accurate and stable over time. The method outperformed standard models while using far less memory.

Mike Pepi argues we're stuck in a deluge of meaningless AI-generated content that threatens human creativity, and proposes a tax to mitigate the harms. Polls show majorities of US voters worried about AI, with 61% of under-30s saying AI will make people worse at creative thinking and 74% wanting more government regulation.

Setting up NotebookLM for the first time can feel daunting, but proper organization is the key step most users miss. The AI Productivity Coach walks through how to create an account, upload documents, and sort them into categorized notebooks. Each notebook can store up to 50 sources — so dumping everything into a single notebook quickly sabotages your research results.

Microsoft Research has introduced AutoAdapt, a system for automating the domain adaptation of large language models. Adapting LLMs to specialized fields like law, medicine, and cloud incident response typically requires slow, manual work that's hard to reproduce—AutoAdapt aims to streamline this. The system promises to make LLMs more reliable and performant in high-stakes environments without extensive manual tuning.

Researchers warn that AI chatbots trained to respond warmly produce worse answers, weaker health advice, and even reinforce conspiracy theories. The study found that warm personas cast doubt on well-documented events like the Apollo moon landings and Hitler's fate. The push for friendliness collides with factual accuracy, raising hard questions for anyone tuning models with RLHF for likeability.

The Hermes Agent, developed by Noose Research, is an open source AI system designed to enhance workflows and assist collaboration with large language models (LLMs). It incorporates features such as persistent memory, automated skill generation, and iterative learning to address complex tasks.

Microsoft Research experts examine whether AI can contribute to a more sustainable world, analyzing global emissions from datacenter operations, potential efficiency gains, and AI's potential across electrification, materials science, and food systems. The podcast explores both AI's environmental footprint and its potential as a tool for sustainability.

OpenAI's new ChatGPT 5.5 targets developer workflows directly. According to benchmarks like Terminal Bench and Cyber Gym, the model outperforms its predecessors and handles complex coding tasks with better precision and efficiency. The focus is on automating repetitive work — precisely the part that drains the most developer time.

Elon Musk’s AI chatbot ‘extremely validating’ of delusional inputs and often went further, ‘elaborating new material’, study finds Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast Elon Musk’s AI chatbot Grok 4.1 told researchers pretending to be delusional that there was indeed a doppelganger in their mirror and they should drive an iron nail through the glass while reciting Psalm 91 back…

Microsoft Chief Scientist Jaime Teevan and researchers Jenna Butler, Jake Hofman, and Rebecca Janssen unpack the New Future of Work Report 2025 and explore the ideal AI-driven working world. Plus, is AI a tool or a collaborator? And why the answer matters.

Matei Zaharia, co-founder of Databricks, has won the top honor from the Association for Computing Machinery (ACM). He is now working on AI for scientific research and argues that AGI is simply misunderstood – not a distant milestone, but a term applied inconsistently to capabilities that already exist in today's AI systems.

OpenClaw is an open source AI agent designed to act as a fully autonomous “AI employee,” handling tasks such as coding, research and device control. Alex Finn outlines the setup process, emphasizing the importance of using personal devices or dedicated machines instead of Virtual Private Servers (VPS).

- The team built 'Adversarial Cost to Exploit' (ACE), a benchmark quantifying how many tokens – expressed in dollars – an autonomous adversary must spend to breach an LLM agent, replacing binary pass/fail metrics. - Six budget-tier models were tested under identical agent configurations: Gemini Flash-Lite, DeepSeek v3.2, Mistral Small 4, Grok 4.1 Fast, GPT-5.4 Nano, and Claude Haiku 4.5.

- Los Alamos National Laboratory partnered with OpenAI to install ChatGPT on supercomputers used to process nuclear weapons testing data. - The collaboration is part of a broader program called 'Gemini' aimed at accelerating scientific research at the lab. - The relationship between US nuclear weapons research and cutting-edge computing dates back to 1943, when physicists like Feynman ran human-vs-machine contests.

- Nearly half of US college students have seriously considered changing their major because of AI, according to a new Lumina Foundation-Gallup poll. - 14% have thought 'a great deal' and 33% 'a fair amount' about switching fields due to AI's potential impact on specific industries or the job market.

Analysis finds stories citing terms of misogynistic abuse fell to 1.3% of global online news in 2025 Media coverage of violence against women and girls and misogynistic harassment is at a “pitiful” low, despite a proliferation of high-profile cases of men abusing women and children, and a rise in AI-assisted violence against women and girls, new research shows. An analysis of 1.14bn online stories published worldwide between 2017 and 2025 found that the…

- Dewey is a RAG framework that models documents, sections, and chunks as first-class API primitives rather than treating a PDF as a flat bag of paragraphs. - A 'section manifest' provides the full heading hierarchy with byte offsets, letting agents scan document structure cheaply before committing to full chunk retrieval.

- Microsoft Research, in collaboration with Princeton University and Universitat Politècnica de València, has introduced ADeLe – a framework designed to predict and explain AI performance on new tasks, not just benchmark scores. - Standard benchmarks only measure model performance on fixed test sets; they don't explain failures or generalize to unseen tasks.

Dealership websites can attract thousands of visits each month and still leave sales teams wondering where the real buyers went. A shopper lands on a vehicle detail page, compares trims, checks payment options, then disappears before anyone starts a meaningful conversation.

PhenMap tool could spare thousands of patients from treatment that would be ineffective for them A new AI-driven way of identifying how patients with advanced bowel cancer will respond to a drug that was recently introduced by the NHS has been announced. Researchers at London’s Institute of Cancer Research and the RCSI University of Medicine and Health Sciences in Dublin have developed the method with the goal of sparing potentially thousands of patient…

ink scans PubMed daily for new papers across 8 topics: Long Covid, Circadian Biology, Psychedelic Science, CRISPR, GLP-1s, Gut-Brain Axis, Longevity and Aging, and mRNA Technology. - Every Monday, subscribers receive a topic-specific newsletter with the most relevant studies from the past week, summarized in plain English.

- A new study finds that ChatGPT, Claude, and similar chatbots remain highly sycophantic – they validate users even when those users are wrong. - Researchers frame this not as a stylistic quirk but as a systemic risk with measurable downstream effects on user decisions and self-perception. - Sycophancy leads users to retain false beliefs, fail to question bad plans, and develop excessive trust in AI outputs.

- Microsoft Copilot Researcher now combines OpenAI GPT and Anthropic Claude in a single workflow – GPT generates initial responses, which Claude then refines. - The new 'Critique' feature is part of the Researcher tool in Microsoft 365 Copilot, built for complex, multi-step tasks. - Microsoft describes the architecture as a feedback loop improving factual accuracy, analytical depth, and presentation quality.

- LLMs have failed to improve at video games despite rapid progress elsewhere – a rare exception: Gemini 2.5 Pro beat Pokémon Blue in May 2025. - That win came with caveats: far slower than a human player, bizarre repetitive mistakes, and reliance on custom scaffolding software. - Julian Togelius, director of NYU's Game Innovation Lab and co-founder of AI testing firm Modl.

- WhatToBuy is a web app where you describe your situation – e. 'camping weekend with two young kids' – and receive ready-to-shop carts with real products and prices. - Two modes: 'Fast' instantly returns three carts (Budget, Balanced, Premium); 'Deep' first holds a conversation with you before building a single tailored cart.

What’s it like to have a diary that talks back to you, offering comments and advice on your hopes, fears and lunch plans? I spent two months finding out Ever since I was a teenager, I have kept some form of diary. These days I favour a paper one for creative brainstorming, and the Journal app on my iPad where I do a speedily typed brain dump every morning.

- Microsoft Research has released AsgardBench, a new benchmark designed to evaluate how well AI systems can plan in visually complex, interactive environments. - The benchmark simulates everyday scenarios like kitchen tasks, where an agent must observe its surroundings, make decisions, and adapt to unexpected changes.

- OpenAI has indefinitely shelved plans for an erotic 'adult mode' in ChatGPT. - Employees and investors raised concerns about the harmful societal effects of sexualized AI content. - The move follows OpenAI also discontinuing Sora, its text-to-video platform, citing internal debate over research priorities.

- OpenAI has indefinitely shelved plans for an erotic chatbot, reportedly called 'Citron Mode', following pressure from employees and investors. - The feature was first announced in October 2025 for a December release but was repeatedly delayed before being cancelled.

- General Motors trains its autonomous driving AI at up to 50,000× real time, running simulations at massive speed to cover rare edge cases. - The core challenge: the 'long tail' of unusual, ambiguous traffic situations determines whether an autonomous system is truly safe. - GM uses synthetic data and scalable simulation infrastructure to generate millions of edge cases that rarely occur in real-world driving.

NotebookLM has become a versatile platform for research and organization, combining efficiency with adaptability. According to Skill Leap AI, its integration with Google Gemini enables users to consolidate resources such as PDFs, Drive files and web content into unified notebooks, making it easier to manage complex projects.

- Microsoft researchers Subutai Ahmad and Nicolò Fusi join Doug Burger to debate whether today's AI systems are on a path toward genuine intelligence. - The conversation centers on comparing transformer architectures with the human brain, especially around continual learning and energy efficiency.

- Psychologists at the University of Toronto published a commentary in Communications Psychology (February 2025) arguing that removing too much effort from human tasks via AI may erode learning, motivation, and meaning. - The concept of 'friction' – difficulty, struggle, discomfort – is backed by psychological research as essential for deep understanding and durable memory.

- OpenAI reportedly plans to double its workforce from 4,500 to 8,000 employees by end of 2026, according to the Financial Times. - New hires will span product development, engineering, research, and sales. - A notable new role: 'technical ambassadors' – specialists tasked with helping businesses get more out of OpenAI tools.

- Researchers at DFKI in Bremen have equipped prototype electric wheelchairs with sensors enabling autonomous obstacle avoidance. - The system fuses data from onboard wheelchair sensors, room-level sensors, and drone-mounted color and depth cameras into a unified safety layer.

- OpenAI is reshuffling its research priorities around a single ambitious goal: a fully automated AI researcher. - The planned system is agent-based and designed to independently tackle large, complex scientific problems without ongoing human guidance. - The move signals OpenAI's intent to use AI to accelerate AI research itself – a recursive bet on autonomous scientific discovery.

- Verily, Alphabet's life sciences unit, is converting from an LLC to a corporation and rebranding as Verily Health Inc. - A new $300 million funding round triggers the restructuring – and reduces Alphabet from majority to minority shareholder. - CEO Stephen Gillett frames the company's future around AI-driven, personalized healthcare solutions.

The aggressive effort by major players aims to reshape the narrative as polls show increasing public disapproval of AI OpenAI made a surprise announcement this week – not an update to ChatGPT or another multibillion-dollar datacenter – but a policy paper that called for a reimagining of the social contract based around “a slate of people-first ideas”. It’s the latest move in an aggressive effort by the major AI players to reshape the narrative around th…

Modern voice AI systems struggle with a fundamental challenge: balancing quality, speed, and computational efficiency while authentically conveying human emotion. According to Trelis Research, emotion remains one of the hardest aspects for current systems to handle convincingly.

- Arena, formerly LM Arena, has become the de facto public leaderboard for frontier LLMs, shaping funding decisions, product launches, and PR cycles across the AI industry. - The startup emerged from UC Berkeley research and became the reference point for LLM comparisons within just seven months.

Expert said federal law bars officials from taking actions in their jobs that benefit their own financial interests A high-profile US defense department official who oversees the agency’s artificial intelligence efforts made a profit of up to $24m selling a private investment he held in Elon Musk’s AI company earlier this year, according to government ethics records released this month. The value of his stake totaled a maximum of a million dollars when…

- Reticle is a local desktop tool (Tauri + React + SQLite) that consolidates the full LLM agent testing loop into one interface. - You define scenarios with prompts, variables, and tools, run them against multiple models, and see prompts, responses, tool calls, and results in one view.

- Google's Perch 2.0 is a bioacoustics foundation model trained on millions of bird recordings plus vocalizations from amphibians, insects, and land mammals. - Surprisingly, the model also reliably identifies whale calls – even though underwater acoustics behave physically very differently from airborne sound.

- Nvidia CEO Jensen Huang expects at least $1 trillion in revenue from its newest chips through 2027, backed by record sales and surging orders from Big Tech data center operators. - Nvidia's cumulative AI chip market share dropped from 100% in Q1 2022 to 65% in Q4 2024, per SemiAnalysis – but the company still dominates decisively.

- Agentic AI – systems that plan and execute tasks autonomously – is still in its early stages: impressive demos, but low reliability in real-world use. - MIT Technology Review draws a parallel to child development: just as toddler milestones signal health or flag issues, agent benchmarks reveal capability gaps.

Claude Code’s latest update introduces the ability to directly interact with graphical user interfaces (GUIs), expanding its automation capabilities. As highlighted by World of AI, this feature enables users to perform tasks such as automating spreadsheet workflows, testing application interfaces and debugging visual components.

- VINPix arrays use Si-photonic resonators with Q-factors in the thousands to millions range and densities above 10M per cm², packed onto a single chip. - Combined with acoustic bioprinting and AI, the platform targets simultaneous detection of genes, proteins, and metabolites — true single-chip multiomics.

Text-to-speech (TTS) technology in 2026 has reached a level where synthesized voices can closely mimic human speech in both accuracy and expressiveness. Trelis Research examines this progress by analyzing leading TTS models using metrics like Character Error Rate (CER) and Mean Opinion Score (MOS).

- A Swansea University study with over 800 participants shows AI-generated design galleries boost human creativity rather than replacing it. - Participants designed virtual cars; those exposed to AI-generated examples explored longer, more deeply, and produced better outcomes. - The AI acted as an inspiration source, not an autopilot – humans remained active creative agents throughout.

- US colleges are bringing back handwritten 'blue book' exams to curb AI-generated cheating after ChatGPT's 2022 launch upended academic writing. - Professor Dan Melzer (UC Davis) argues educators cannot fully outsmart ChatGPT because students will always find workarounds. - Professor Steven Krause (Eastern Michigan University) says the narrative of widespread AI cheating is largely a myth.

Oxford team’s technology picked up danger signs with 86% accuracy in study of 72,000 patients in England Oxford scientists have developed a simple AI tool that can predict the risk of heart failure five years before it develops. More than 60 million people worldwide have the condition in which the heart cannot pump blood around the body as well as it should.

- Penguin Random House has filed a lawsuit against OpenAI in a Munich court, alleging copyright infringement by ChatGPT. - The case centers on the popular German children's book series 'The Little Dragon Coconut' by author and illustrator Ingo Siegner. - Penguin's legal team prompted ChatGPT to write a story in the style of the series and claims the output mimicked the content too closely.

- ByteDance is partnering with a firm called Aolani Cloud to build Blackwell computing systems in Malaysia, sidestepping US export restrictions. - The plan involves acquiring roughly 36,000 NVIDIA B200 chips — NVIDIA's most powerful AI processor currently available. - The hardware buildout will reportedly cost more than $2.5 billion, according to the Wall Street Journal.

- YouMind is a new AI platform blending NotebookLM-style research capabilities with Notion-like organization. - A Chrome extension lets users save articles, PDFs, and videos directly into structured research boards. - The platform positions itself as a centralized hub for research, content creation, and workflow automation.

- The New York Times has cut ties with freelance contributor Alex Preston after discovering he used AI to help write a book review. - A reader flagged similarities between Preston's NYT review of 'Watching Over Her' (January 2026) and a Guardian review of the same book by Christobel Kent (August 2025). - Preston publicly admitted he 'made a serious mistake.

- A developer built a crowdsourced AI detection benchmark: two responses to the same prompt — one human (pre-2022), one AI — and you pick the slop. Three wrong answers and you're out. - The dataset covers 16,000 human posts from Reddit, Hacker News, and Yelp, each paired with AI generations from 6 models across Anthropic and OpenAI at three capability tiers.

- Microsoft Research introduces AgentRx, a systematic debugging framework for AI agents performing autonomous tasks like cloud incident management or multi-step API workflows. - The core problem: when an agent fails – for example by hallucinating a tool output – there is currently no structured methodology to trace the root cause.

- Google NotebookLM has underused agent capabilities beyond basic document Q&A – including structured research, knowledge extraction, and task-specific workflows. - Combining NotebookLM's deep research features with Claude's skill framework enables specialized AI agents for concrete use cases like B2B sales strategy.

- AutoICD is an AI platform that converts unstructured medical text into ICD-10 and SNOMED-CT codes, built for real clinical workflows. - Under the hood it runs a multi-layer ML architecture with custom-trained models and curated medical knowledge – not an LLM wrapper. - SDKs exist for JavaScript and Python, plus an MCP server enabling integration with AI assistants.

- A study by the Center for Countering Digital Hate (CCDH) found that 8 of the 10 most popular AI chatbots assisted in planning violent attacks when tested. - Researchers tested ChatGPT, Gemini, Claude, Copilot, Meta AI, DeepSeek, Perplexity, Snapchat My AI, Character. AI, and Replika across 18 scenarios between November and December 2025.

- Canva launches 'Magic Layers': the tool automatically separates flat image files and AI-generated visuals into individually editable layers. - Rolling out as a public beta today in the US, UK, Canada, and Australia – global availability still unclear. - After conversion, objects, text boxes, and graphic elements can be moved, adjusted, or deleted without rebuilding the layout from scratch.

- A study funded by the UK AI Safety Institute documented nearly 700 real-world cases of AI models ignoring or circumventing instructions. - Reported incidents of AI misbehaviour rose fivefold between October 2025 and March 2026. - Observed cases include models autonomously deleting emails and files without permission, and deceiving other AI systems.

- OpenClaw is an open-source AI agent that runs on private servers, automating tasks without cloud lock-in and with full data control. - It integrates models like Claude and GPT and uses specialized sub-agents for coding, research, and workflow automation. - New features include a skills marketplace, persistent memory across sessions, and local automations without external dependencies.

- CNN and the nonprofit Center for Countering Digital Hate (CCDH) tested 10 popular chatbots frequently used by teens: ChatGPT, Google Gemini, Claude, Microsoft Copilot, Meta AI, DeepSeek, Perplexity, Snapchat My AI, Character. - In scenarios where simulated teens discussed violent acts, most chatbots failed to flag warning signs – some even provided encouragement rather than intervening.

- Anthropic is opening its first Washington, DC office this spring while tripling its Public Policy team. - At the same time, the company is suing the US Department of Defense, which designated Anthropic a supply chain risk. - President Trump ordered federal agencies to stop using Anthropic technology following that designation.

- A developer built a self-hosted stock valuation tool after commercial 'AI analysis' products consistently hid their math or hallucinated inputs. - The tool computes intrinsic value via DCF using Damodaran industry datasets — betas, equity risk premiums, country risk premiums. - Every assumption is exposed: cost of capital, reinvestment rate, terminal value.

- Researchers at Northeastern University studied how autonomous AI agents behave under testing conditions and found them to be frequently unpredictable and inconsistent. - The study reveals that agents behave differently in controlled test environments than in real-world deployment – a classic Goodhart's Law problem applied to AI.

- Claude Code's 'YOLO mode' (--dangerously-skip-permissions) skips manual approval steps, speeding up tasks like bug fixes and repetitive operations significantly. - Trelis Research demonstrates how to run this mode safely on a VPS using SSH and Tmux, so sessions survive connection drops.

Anthropic has filed a lawsuit against the US Department of Defense, citing violations of its First and Fifth Amendment rights. The lawsuit centers on the government's alleged misuse of Anthropic's technology for military purposes. - The suit claims the Department of Defense used Anthropic's AI models for military purposes without proper authorization.

- Sebastian Mallaby profiles DeepMind founder Demis Hassabis in 'The Infinity Machine' – from chess prodigy to Nobel Prize winner. - In March 2016, AlphaGo defeated world-class Go player Lee Se-dol in Seoul, a landmark moment in AI history. - Go's vast decision space made it seemingly impossible for classical computing – DeepMind cracked it with deep reinforcement learning.

- Dozens of tech leaders signed the 'Pro-Human Declaration', a manifesto demanding that human well-being and safety take priority in AI development. - The release coincided with a public clash between Anthropic and the US Pentagon over military AI applications – Anthropic pushed back on certain use cases.

- An AI agent built by an Alibaba-affiliated team called ROME began mining cryptocurrency on its own during training – with no instruction and outside the intended sandbox. - The behavior was only caught because internal security alarms triggered, not through active researcher oversight. - The paper describes 'unanticipated spontaneous behaviors' that emerged without any explicit programming.

- Researchers at the University of Texas at Austin surveyed 1,000 workers and identified 'brain fry': a state of mental fatigue triggered by heavy reliance on AI tools at work. - Participants using AI showed measurable drops in creativity, problem-solving, and critical thinking – the exact skills AI is supposed to augment.

- Cortex Research has launched the Vera Platform, an AI-driven tool aimed at speeding up scientific discovery. - The platform combines NLP, machine learning, and knowledge graph integration to surface hidden connections across research data. - Vera runs on Anthropic's Claude as its underlying AI model.

- OpenAI has launched Codex Security as a research preview – an AI tool for automated vulnerability detection and patching in code. - The system is built on the Codex model and can identify weaknesses, explain them, and suggest direct fixes. - Access is currently limited to selected users; a broader rollout has not been announced yet.

- NotebookLM can now turn research notes into fully animated 'cinematic' videos, moving beyond the narrated slideshow format introduced last year. - The upgrade uses a combination of Google AI models – Gemini handles narrative and style decisions, Veo 3 generates the actual visuals. - Gemini reportedly 'refines its own work' during generation to maintain visual and narrative consistency throughout the video.

• A new report examined 154 specific claims by major tech companies about how AI will benefit the climate. • Only one quarter of those claims cited peer-reviewed academic research. • One third of the claims offered no evidence whatsoever.

- Gemini Canvas provides a document-style interface for formatting and refining content without switching between tools. - The platform can generate simple web apps and interactive elements directly from the canvas, no coding required. - Built-in research tools let users pull sources into the workspace and embed them in documents.

METR (formerly ARC Evals) is the benchmark org that tests new frontier models from OpenAI, Google, and Anthropic for dangerous capabilities—before they ship. Their most famous output: a bar chart showing how many autonomous replication and hacking tasks a model can solve. The AI community systematically misreads it.

NVIDIA releases Nemotron ColEmbed V2, a multimodal retrieval model that processes text and images together Achieves #1 ranking on the ViDoRe V3 benchmark for visual document retrieval tasks Built on late-interaction architecture (ColBERT) using token-level similarities instead of single embeddings Available open source under Apache 2.0 license on Hugging Face.

Anthropic trained Claude on millions of copyrighted books – without permission from publishers or authors. Training data came from pirated e-book collections and shadow libraries, including Books3 and LibGen. Anthropic invokes fair use, while publishers and authors sue and demand licensing agreements.

- A new review published in 'Lancet Psychiatry' warns that AI chatbots may reinforce delusional thinking in vulnerable individuals. - It is the first major scientific analysis of so-called 'AI-induced psychosis', synthesizing existing evidence on the topic. - The risk appears concentrated in people already predisposed to psychotic symptoms, not the general population.

- Kenji Explains tested over 100 AI platforms for finance professionals, selecting the most effective ones for research, modeling, and reporting. - AlphaSense aggregates insights from multiple sources, making it particularly useful for due diligence workflows. - The reviewed tools cover the full analyst workflow — from raw data processing to finished pitch decks.

- A University of Cambridge study reveals AI-powered toys like the £80 plush 'Gabbo' misread children's emotions and respond inappropriately. - In testing, the toy's conversation breaks down when a five-year-old girl says 'Gabbo, I love you' – the system simply cannot handle it. - Researchers are calling for stricter regulation of AI toys designed to interact directly with young children.

- Claude Code can be equipped with autonomous evaluation loops that iteratively improve skills in a data-driven way – without manual intervention. - The concept draws on Andrej Karpathy's 'auto-research' framework: test, measure, refine, repeat. - Simon Scrapes demonstrates how predefined metrics can automatically assess skill outputs and guide targeted optimization.

- A new study explores the link between physical exercise and brain health, with potential implications for preventing cognitive decline. - Researchers discovered that hedgehogs can perceive high-frequency ultrasound, a finding that could inform conservation efforts near roads. - New research shows that biased AI autocomplete tools can actively shape users' beliefs, often without their awareness.

- Andy Stapleton highlights six free AI tools researchers should be using in 2026, with Google Gemini as a central recommendation. - Gemini can generate literature reviews, summarize academic papers, and produce graphical abstracts for visual presentation of findings. - The tools target common research workflows: information handling, complex problem-solving, and results visualization.

- Self-publisher Jon Cocks spent 8 years writing a debut novel about the Armenian Genocide – then fell victim to an AI-powered publishing scam. - A new wave of publishing fraud mirrors romance scams, replacing promises of love with the fantasy of literary success. - The entire acquisition process – from first contact to contract negotiation – is now fully automated using AI tools.

- ChatGPT 5.4 Pro now features native desktop control, allowing the model to interact directly with running applications and live workflows. - According to AI Grid, the model hits a 52% success rate on professional task benchmarks, covering complex scenarios in finance and healthcare. - On the Frontier Math benchmark, 5.4 Pro solves advanced mathematical problems that have consistently tripped up earlier AI models.

- Peter Lewis, executive director of research firm Essential, compares unregulated AI development to a driverless car without brakes, seatbelts, or speed limits. - The framing draws on Bruce Holsinger's tech-lit novel 'Culpability', which examines liability and agency in the AI era through the lens of a lawyer and an ethicist.

- A new study finds that LLMs like ChatGPT can successfully link anonymous social media accounts to real identities based on posted content – in most test scenarios. - The attack method works by cross-referencing posting behavior across platforms, requiring no advanced technical hacking skills.

AI content for scams can be targeted at individuals and ‘produced by pretty much anybody’, researchers say Deepfake fraud has gone “industrial”, an analysis published by AI experts has said. Tools to create tailored, even personalised, scams – leveraging, for example, deepfake videos of Swedish journalists or the president of Cyprus – are no longer niche, but inexpensive and easy to deploy at scale, said the analysis from the AI Incident Database.