AI News Archive: March 2026 - Monthly Highlights
100 articles published this month
AI productivity gap: top agent beats baseline in 1 of 15 runs, 26.5% subtasks
Why does the hype around AI productivity often feel out of step with what actually gets delivered?
Meta launches prescription-optimized smart glasses amid privacy criticism
Why does Meta’s newest hardware matter now? The company just rolled out a version of its Ray‑Ban‑styled glasses that can accommodate prescription...
AI splits art schools as program links students with Adobe, Google
AI is reshaping art education, and not everyone agrees on the right path. Some programs have begun wiring classrooms directly to the tech giants that...
Anthropic's Claude Code source leak prompts guidance for users
Anthropic’s Claude Code has been thrust into the spotlight after its source code appeared online, prompting a flurry of concern across developer...
Anthropic finds LLMs could theoretically handle 80% of tasks across many jobs
Anthropic’s latest internal analysis has sparked a quiet stir among analysts watching the intersection of large‑language models and the labor market.
Zero‑Budget Full‑Stack App Uses Free Whisper, GLM‑4.7‑Flash and FastAPI
Building a complete web application without spending a dime sounds more like a thought experiment than a practical guide, yet the project outlined in...
Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Nvidia’s latest software rollout nudges its deep‑learning super‑sampling tech into a new performance tier.
Qwen3.5-Omni writes code from spoken instructions, fixes voice token lag
Why does real‑time speech still sound jittery? While most models can generate text fluently, the audio stream often lags behind, dropping words or...
OpenAI launches Codex plugin for Claude Code with local settings and Review Gate
OpenAI’s newest offering plugs its Codex engine directly into Anthropic’s Claude Code, turning the latter into a front‑end for an already‑installed...
Softr unveils AI-native platform, calling itself the ‘Canva for web apps’
Softr’s latest rollout promises a new way for teams without coding chops to piece together functional web applications.
Nvidia-backed ThinkLabs raises USD 28M, shows AI power‑grid results with SCE
ThinkLabs, the AI‑focused startup backed by Nvidia, just closed a $28 million funding round aimed at easing the strain on America’s aging power...
Microsoft adds Critique and Council to Copilot following USD 1B Disney blindside
Microsoft just turned its Copilot Researcher into a two‑engine engine, rolling out the Critique and Council features in a single update.
Midjourney engineer releases open‑source Pretext to curb browser layout stalls
Why do modern web pages sometimes feel sluggish when you scroll or when on‑the‑fly text appears?
MetaClaw trains AI agents via Google Calendar, turning failures into rules
MetaClaw promises to keep AI assistants learning even when you’re busy in a conference call.
Insiders call AI the 'Ozempic' of music as over half of hip‑hop is AI‑generated
Why does this matter? Because producers are swapping out dusty vinyl crates for algorithms, and the ripple is already reshaping the sound of a genre...
AI sycophancy cuts apologies, raises double‑downs; lifts moral trust
Why does it matter when a chatbot mirrors the tone you expect? Researchers set out to see whether an AI’s conversational style could sway how people...
Altman, Science VP hype AI‑assisted dog cancer vaccine despite lacking proof
Why does a Silicon Valley headline about an AI‑driven dog vaccine matter to a pet owner in Sydney?
AI models fabricate image descriptions; benchmarks miss the shortcuts
Why does it matter when a system tells you it “sees” something it never actually looked at?
Cohere's open-weight ASR model reaches 5.4% WER, ready for production use
Cohere’s newest speech‑to‑text system hits a 5.4 % word error rate, a figure that sits at the low end of what many enterprises consider acceptable...
Anthropic unveils secret 'Mythos' model for AI development in internal testing
Anthropic has slipped a new language model into its labs, calling it Mythos. The project stays under wraps, but insiders say it powers a string of...
Deezer's AI tool targets royalty theft from mass‑produced AI music
Why should anyone care about a streaming service’s new tech? Because the flood of AI‑generated tracks is already testing the limits of existing...
Amazon’s Spring Sale Offers AI‑Powered Bird Feeder for Migration Season
Spring is in the air, and the annual bird migration is already filling backyards with flocks of warblers, finches and swallows.
Kuo Zhang says work is moving from step management to outcome‑driven AI agents
Kuo Zhang, a senior executive at Alibaba.com, sat down for an exclusive Q&A to explain what the company calls “agentic futures.” In the conversation,...
Suno launches v5.5, AI music model lets users train on their own voice
Suno’s new v5.5 update pushes the platform a step further into personalization. While earlier versions let users generate melodies and backing...
Samsung's TikTok videos lack AI disclosures despite YouTube AI tags
Why does a tech giant’s presence on a short‑form platform matter when the rules governing AI‑generated content are still being tested?
Disney pledged USD 1 billion equity stake in OpenAI and to become a major customer
Disney’s move into generative‑AI has been anything but subtle. The company has been courting OpenAI’s technology for months, positioning the...
Bluetti Sora 500’s fabric backing and kickstand system differs from Zoupw 480W
Bluetti’s Sora 500 has been marketed as a compact, high‑output solar panel, promising more watts per square foot than many rivals.
Judge says Trump had no authority to blacklist Anthropic, War Dept label
A federal judge has ruled that former President Donald Trump never possessed the statutory power to place AI startup Anthropic on a government...
Company unveils 5-point Community-First AI plan to curb data center energy use
Why does a tech giant’s internal roadmap matter to anyone outside its own walls? Because data centers now sit at the crossroads of AI ambition and...
IndexCache sparse attention optimizer makes long-context AI 1.82× faster
IndexCache promises to make long‑context language models noticeably quicker. The new sparse‑attention optimizer, announced under the headline...
Free API that evolved from slow web search to top AI tool, beyond scraping
The roundup of free web APIs just added a service that’s quietly reshaped how autonomous agents pull data from the internet.
Meta unveils open-source brain AI, adds Scrunch site audit and Suno v5.5
Meta’s latest push into open‑source AI isn’t just another research paper; it’s a bundle of tools aimed at developers and marketers alike.
Partnership on AI names BBC R&D, Capital One execs as board chair, vice chairs
The Partnership on AI has reshuffled its top‑level governance, tapping senior figures from three distinct sectors to steer its next phase.
AI assurance experts meet to build infrastructure for safe, high‑quality systems
The AI community has been wrestling with a simple question: how do we move from flashy prototypes to systems that people can actually rely on?
Judge grants Anthropic temporary block on Pentagon ban in federal court
A federal judge has issued a temporary injunction that halts the Pentagon’s recent prohibition on using Anthropic’s artificial‑intelligence tools.
David Sacks removed as White House AI and crypto czar after Trump fundraiser
David Sacks is out. The White House announced Tuesday that the former AI and crypto czar will no longer serve in the role that gave him direct access...
Apple's Siri to support third‑party AI chatbots via new 'Extensions' feature
Why does this matter now? Siri has long been a closed voice assistant, answering queries with Apple‑built models while third‑party bots linger on the...
Kensho builds a LangGraph multi‑agent framework for trusted financial data
Financial analysts at S&P Global have long wrestled with a core problem: pulling reliable numbers from dozens of internal repositories without...
Import AI memories and chat history to Gemini for seamless continuity
Why does continuity matter when you switch AI assistants? Most users spend weeks—or months—teaching a chatbot the details that make conversations...
Google expands Search Live to dozens of languages with Gemini 3.1 audio
Why does this matter now? Google just announced that its Search Live feature will support “dozens of languages,” a move that pushes the...
Study finds overly flattering AI advice can impair users' judgment
Why does it matter when a chatbot tells you what you want to hear? A new study, published under the title “Sycophantic AI can undermine human...
Meta readies Oakley AI glasses and Ray‑Ban Display glasses monocular display
Meta’s push into wearable AI is reaching a new milestone. After a year in which its smart‑glass lineup saw sales more than triple, the company is...
Enterprises with billions in infrastructure want AI to integrate data and APIs
Enterprises that have poured billions into legacy data centers and custom‑built APIs now face a stark dilemma: their massive, aging stacks are...
MedGemma Impact Challenge winners use AI to convert local notes into WHO data
The MedGemma Impact Challenge just crowned its first cohort of innovators, and their prize isn’t a trophy—it’s a prototype that could reshape disease...
Intercom's Fin Apex 1.0 beats GPT‑5.4, Claude Sonnet 4.6 in service
Intercom just rolled out Fin Apex 1.0, a post‑trained large language model built for customer‑service tasks.
Build a Smolagent with Qwen2.5-Coder-32B-Instruct using Hugging Face token
If you’ve ever wanted a tiny, purpose‑built AI that can write or debug code without pulling in a heavyweight framework, the “smolagent” concept...
Wikipedia bans AI-generated articles, allows AI only for copy editing or translations
Why has the world’s largest free encyclopedia suddenly tightened the reins on artificial‑intelligence tools?
Gemini 3.1 Flash Live cuts latency, boosts pitch and pace detection
Gemini 3.1 Flash Live arrives as the next step in Google’s push for real‑time AI chat.
Middleware Enables Deep Customization of Agent Core Loop and Harness
Developers building on LangChain quickly discover that the out‑of‑the‑box harness covers most use cases, but it can feel limiting when a project...
Webtoon adds AI localization with human QA review for error reports
Webtoon is rolling out AI‑driven localization across its comics platform, promising faster translations for its global audience.
Mistral AI releases TTS model that beats ElevenLabs, runs on old chips, supports 9 languages
Mistral AI just dropped a new text‑to‑speech model that it claims outperforms ElevenLabs on every benchmark it’s been tested against.
Anthropic ships Claude with Mac control, joins OpenClaw, NemoClaw race
Anthropic just added Mac‑control to Claude, turning a conversational model into a hands‑on assistant that can click, type and launch apps on a...
Tely AI offers hands‑off recommendations in ChatGPT, Google, Claude in a week
Businesses are feeling the pressure of a new kind of visibility race—one that pits their brand against the answers generated by AI chatbots.
Uber partners with Pony AI and Verne to launch Europe's first robotaxi service
Europe has never seen a robotaxi service on the scale Uber envisions. The company’s push into autonomous mobility comes at a moment when cities are...
Spotify's Artist Profile Protection: artists approve releases to block AI fakes
Why are musicians suddenly checking a new box before their songs appear on streaming shelves?
Disney distances from OpenAI as metaverse and AI bets continue to falter
Disney has been trying to juggle two high‑profile experiments that have yet to deliver the promised returns.
Oracle unifies AI data stack to address enterprise agents' data-tier failures
Oracle is stitching together the pieces of its massive database empire to give AI‑powered assistants a cleaner line of sight.
Google's TurboQuant boosts AI memory 8×, slashes serving costs half
Google unveiled TurboQuant this week, an algorithm that claims to multiply AI‑model memory bandwidth by eight while halving the expense of serving...
Google's TurboQuant cuts AI key‑value cache size without quality loss
Google’s latest tweak to its language‑model stack promises to shrink the memory footprint of large‑scale inference without denting output quality.
xMemory reduces token usage and context bloat versus MemGPT's raw logging
Early AI agents often treat every exchange as a line in a ledger, appending each utterance to a growing transcript.
Vibe Creates Private AI Financial Analyst: Data stays local, no API fees
Vibe’s latest project tackles a niche that many developers overlook: building an AI‑driven financial analyst that runs entirely on‑premises.
Lyria 3 supports image‑to‑music input, shaping audio in Google AI Studio
Google’s latest foray into generative sound arrives with a model that pushes past the usual text prompts.
5 Ways to Collaborate with Agentic Advisors to Keep AI Tools Ahead
Marketing moves fast, and your AI tools should do more than just keep up—they need to push you ahead.
Senate Democrats seek to codify Anthropic's red lines on weapons, surveillance
Senate Democrats are pushing a bill that would lock Anthropic’s self‑imposed limits on autonomous weapons and mass‑surveillance tech into law.
Disney ends USD 1 billion OpenAI deal as Sora shutdown plans
Disney’s decision to walk away from a $1 billion partnership with OpenAI has sent ripples through both the entertainment and AI sectors.
Anthropic introduces safer auto mode for Claude Code, balancing handholding and autonomy
Anthropic is nudging its coding assistant toward a more measured kind of independence.
Samsung Galaxy A57 gains IP68 rating, slimmer bezel, higher price
Samsung’s latest mid‑range offering arrives with a price tag that nudges it closer to flagship territory, prompting a quick look at what’s actually...
OpenAI's Sora cut; Airia's 2026 AI Report details enterprise adoption trends
OpenAI just pulled the plug on Sora, a move that sent ripples through the startup community and left many wondering how quickly even the biggest labs...
Mozilla dev launches cq, a Stack Overflow‑style hub for agents
Mozilla’s latest open‑source effort, cq, aims to give autonomous agents a place to post solutions and borrow tricks the way developers turn to Stack...
OpenAI to discontinue Sora video generator after Disney megadeal fizzles
OpenAI’s foray into AI‑generated video seemed poised to reshape content creation.
OpenAI shuts down Sora AI video model after Disney's USD 1B investment
OpenAI announced today that it is pulling the plug on Sora, its AI‑driven video generation model, and the associated app and API.
Arm's first AGI CPU, up to 136 cores, to power Meta AI datacenters this year
The tech world has been watching Arm’s push into high‑density AI silicon for months.
Apple tests standalone Siri app ahead of iOS 27, macOS 27 debut at WWDC 2026
Apple is quietly trialing a new, self‑contained Siri app, a move that hints at a broader redesign of the voice assistant ahead of next year’s...
Cloudflare launches Workers; Rivet’s Secure Exec supports Vercel, Kubernetes
Cloudflare just rolled out a new class of Dynamic Workers that claim to run AI‑agent code without containers and at speeds up to a hundred times...
Google adds Gemini checkout partners while OpenAI upgrades ChatGPT product views
Google is nudging its Gemini model into the checkout lane, lining up new retail partners to surface product picks directly inside a chat.
ChatLLM launches Route LLM to auto‑select models for writing, code, research
Juggling separate AI services for drafting prose, debugging code, or digging up citations has become a routine headache for many professionals.
Moda assesses migration of Design Agent from custom LangGraph loop
Moda’s engineering team has been busy turning experimental AI prototypes into tools that can actually ship.
Google TV adds three Gemini features for interactive, guided walkthroughs
Google TV is rolling out three new Gemini tools that turn the living‑room screen into a kind of digital classroom.
Liquid‑cooled AI systems make storage an active cooling and GPU partner
Why does this matter now? As AI models swell and GPU farms push power envelopes, designers are turning to liquid‑cooled chassis to keep chips from...
Anthropic releases Claude Code, Cowork for macOS in research preview
Anthropic just rolled out a new research preview that lets its Claude models interact directly with a computer.
10 X Accounts for LLM Updates, Including the ‘Largest AI Newsletter’
Keeping up with large‑language‑model news feels like chasing a moving target. Every day a new model drops, a startup announces funding, or a research...
How to Build Custom Codex Skills for Reusable Coding Workflows
Developers who spend hours tweaking the same snippets often wonder if there’s a shortcut.
Nanobot guide shows how to configure OpenAI GPT‑5.3‑Codex agent for WhatsApp
Getting a nanobot up and running on WhatsApp isn’t just about flipping a switch; it hinges on a handful of settings that tell the agent where to pull...
Zuckerberg creates personal 'CEO agent' as Anthropic's Claude gets remote control
Why does this matter now? Because the line between executive decision‑making and algorithmic assistance is blurring faster than most boardrooms...
DeerFlow 2.0 Sets Defined Architecture and Scoped Tasks for Local AI Agents
Local AI agents have been sprouting across open‑source repos, promising everything from autonomous research assistants to creative bots.
Nvidia's 3B Nemotron-Cascade 2 wins math and coding gold; recipe open‑source
Nvidia’s latest 3‑billion‑parameter model, Nemotron‑Cascade 2, just swept the top spots in both math and coding benchmarks, earning gold medals that...
LangChain to Appear at Google Cloud Next 2026 with Atlassian and Google Leaders
LangChain is set to take the stage at Google Cloud Next 2026, joining forces with Atlassian and Google to showcase how open‑source tooling can fit...
Nvidia CEO clarifies DLSS 5 concerns in lengthy Lex Fridman interview
In a nearly two‑hour conversation on the Lex Fridman Podcast, Nvidia’s chief executive Jensen Huang tackled the backlash that followed the rollout of...
Apply: AI agents that reason, plan, execute workflows at Transform 2026
Enterprises are racing to turn autonomous software into a practical advantage, yet few teams have actually delivered agents that move beyond...
Luma AI's Uni-1 beats Google and OpenAI, costs up to 30% less
Luma AI’s latest release, Uni‑1, has already drawn attention for two stark claims: it outperforms the image‑generation models from Google and OpenAI,...
Cursor's Composer 2 built on Chinese AI model, exposing Western open-source gaps
Cursor’s latest release, Composer 2, quietly runs on a Chinese‑origin model, a fact that slipped past most reviewers until a deep dive of the code...
Teens await sentencing for AI‑generated nude images as parents sue school
The courtroom is waiting, but the stakes stretch far beyond any single verdict. Two teenagers stand accused of using generative AI to produce...
Google NewFront 2026: Gemini advantage shows integrated G‑MP campaigns add value
Google’s NewFront this year put the spotlight on Gemini, its newest generative‑AI engine, and how it plugs into the broader Google Marketing Platform...
AI tools aid lawyers as courts limit expert reports, say practitioners
Courts across the UK are tightening the gate on expert testimony, leaving many inquests without the specialist insight families rely on.
Samsung adds AirDrop to Quick Share on Galaxy S26, launches in Korea
Why does this matter? For years, Android users have watched iPhone owners swap files with a single tap, while Samsung’s own Quick Share required a...
Elon Musk's 'Terafab' AI Chip Factory Featured in Quick Hits of New AI Tools
Why does a chip factory built by Elon Musk matter to the tools your team will actually use tomorrow?
New self-summarization RL technique trims action history to 1,000 tokens
Why does trimming a model’s memory matter? In reinforcement‑learning setups where a language model continuously writes, the action log can balloon...
AI Influencer Awards to Celebrate Creative Talent Behind Bots, Offer Prize Fund
AI’s newest celebrities are about to get a spotlight of their own. A fresh contest is rolling out this spring, aimed squarely at the people who...
Human-in-the-Loop: Training Wheel Mode Lets Agents Prove Themselves in Risky Ops
When autonomous systems start tackling tasks that could affect safety or finances, the margin for error shrinks dramatically.
Developers say AI‑generated games feel unlike human‑made; audiences don't connect
At the industry’s biggest developer showcase this year, every booth was flashing AI demos, yet the actual game line‑up remained conspicuously...