AI News
AI news that matters. Updated daily.
No stories match your filters.
Pentagon Labels Anthropic a 'Supply Chain Risk,' Orders Military to Drop Claude
The standoff between Anthropic and the U.S. government just escalated to a level nobody in AI predicted a year ago.
Digg Shuts Down Two Months After Relaunch as AI Bots Overwhelm the Platform
Two months. That's how long Digg lasted after Kevin Rose and Alexis Ohanian relaunched the once-iconic link aggregator in January 2026.
Critical 9.8 Vulnerability in ModelScope Lets Attackers Hijack AI Agents
A vulnerability rated 9.8 out of 10 on the CVSS severity scale has been disclosed in ModelScope's MS-Agent framework, and it demonstrates exactly the kind of AI agent security nightmare that researchers have been warning about.
AI Shopping Agents Still Can't Buy Anything - Here's What's Blocking Them
Only 16% of US consumers trust AI to make payments on their behalf. That number alone explains why the "agentic commerce" future everyone keeps predicting remains stuck in pilot programs and demos.
Developers Are Using Redis to Coordinate Multiple AI Coding Agents
Getting multiple AI coding agents to work together usually means stitching together MCP servers, custom APIs, or shared file systems with careful orchestration logic. A simpler pattern is gaining traction among developers: just use Redis.
Developer Uses OpenAI Codex to Reverse Engineer a 1975 FORTRAN Compiler
A 50-year-old FORTRAN compiler, written for a processor most people have never heard of, just got a thorough autopsy from an AI.
Nyne Raises $5.3M to Feed AI Agents the Context They Lack
A father-son team just raised $5.3 million to solve one of the more stubborn problems with AI agents: they don't know enough about the people they're supposed to help.
Grammarly Pulls 'Expert Review' Feature After Writers Revolt and Lawsuit
"You rapacious information and identity thieves better get ready for me to go full McConaughey on you. Also, you suck."
Microsoft Bringing Copilot AI Assistant to Xbox Consoles in 2026
Microsoft is putting an AI assistant on your Xbox. Sonali Yadav, Xbox's product manager for gaming AI, confirmed during a panel at the Game Developers Conference (GDC) that the company's Gaming Copilot will arrive on "current-generation consoles" this year and expand to "more services" beyond the console itself.
A Software Engineering VP Argues Agile Needs Rewriting for the AI Era
"AI made producing software cheap, but understanding it is still expensive." That one sentence, from a 20-year engineering veteran, captures a tension every team using AI coding tools is feeling right now.
Smellcheck: Open-Source CLI Detects AI-Written Text Without Any API Calls
A new open-source CLI tool called Smellcheck takes a refreshingly low-tech approach to detecting AI-generated text. Instead of running content through a machine learning classifier or sending it to a detection API, it scans for four categories of telltale patterns that LLMs leave behind like fingerprints.
Argo Turns Playwright Test Scripts Into Demo Videos With AI Voiceover
Developers who already write Playwright browser tests can now repurpose those scripts as product demo videos. Argo, a new open-source CLI tool, records browser sessions driven by Playwright, overlays UI highlights, and generates voiceover from a simple JSON manifest.
GlassWorm Attack Hides Malicious Code in 151+ GitHub Repos Using Invisible Unicode
151 GitHub repositories. That's how many were compromised in a single week by an attack technique most developers have never heard of, and literally cannot see.
Engineer Uses ChatGPT to Help Design a Cancer Vaccine for His Dog
When Paul Conyngham's rescue dog Rosie was diagnosed with advanced mast cell cancer in 2024, vets gave her one to six months to live. Conyngham, a Sydney-based AI consultant and Director of the Data Science and AI Association of Australia, decided to use the tools he knew best.
Claude Now Generates Weather Widgets, Charts, and Recipe Cards Inline
Anthropic has quietly rolled out a set of visual and interactive response types in Claude that go well beyond text. Instead of describing a weather forecast or a recipe in paragraph form, Claude now renders actual widgets: weather cards with forecasts, recipe cards with adjustable servings and measurement toggles, and custom charts built from uploaded data.
Runnrr: A Free Kanban Board Built to Feed Context Into Claude Code
One persistent annoyance with AI coding assistants: your session ends, and all that accumulated context vanishes. Runnrr, a new free tool from developer Dmitrijs Vitjazevs, tries to solve this with a dead-simple Kanban board designed specifically for Claude Code users.
Monet Brings Grid-Based Session Management to Claude Code Power Users
Running one Claude Code session is straightforward. Running a dozen in parallel across different parts of a codebase? That gets messy fast. Monet is a new Mac app built specifically for developers juggling multiple Claude Code agents at once.
Promptctl Pipes Local LLM Prompts Into Remote SSH Sessions
Running AI tools on remote servers usually means one of two bad options: give an LLM agent direct SSH access to your infrastructure, or install AI tooling on every server you manage. Promptctl takes a third path.
Context Gateway: Open-Source Proxy Compresses AI Agent Context to Cut Costs
Anyone who has watched Claude Code or a similar coding agent burn through tokens reading files knows the problem: a single grep or file read can dump thousands of irrelevant lines into the context window (the amount of text the AI model can "see" at once). Most of that content is noise the model never needed.
Stint Automates Multi-Agent Claude Workflows With Git Branch Isolation
Most AI coding tools still expect you to sit there, prompting one task at a time. Stint takes a different approach: queue up a list of development goals, walk away, and let multiple Claude instances work in parallel across isolated git branches.
For Blind Users, AI Tools Have Become Essential Accessibility Tech
Three years ago, a blind person who received an inaccessible PDF had limited options: ask someone sighted for help, run it through patchy OCR, or skip it entirely. Now they paste it into Claude or ChatGPT and get a readable version in seconds.
Claude's 1 Million Token Context Window Goes GA at Standard Pricing
Anthropic just made the 1 million token context window generally available for both Claude Opus 4.6 and Claude Sonnet 4.6. The feature had been in beta, but as of March 13, it works out of the box with no special headers, no waitlist, and no price premium.
AI Coding Benchmarks Miss a 2x Quality Gap Between Models
Three AI coding models score within one percentage point of each other on test pass rates. But when you actually look at the code they write, one produces solutions nearly twice as close to what a human developer would ship.
Job Posting Data Shows Claude Catching Up to ChatGPT in Enterprise
Last year, saying "we use AI" in a job post meant ChatGPT. That default is shifting.
Reports Suggest Claude Code Users May Hit A/B Tests That Affect Quality
If Claude Code has felt inconsistent lately, you might not be imagining it.
ByteDance Reportedly Gains Access to Top-Tier Nvidia AI Chips
ByteDance, the Chinese parent company of TikTok, has reportedly gained access to Nvidia's top-tier AI chips, according to the Wall Street Journal. The development is notable given the US government's ongoing efforts to restrict advanced semiconductor exports to China.
Anthropic Drops Long-Context Pricing: 1M Tokens Now at Standard Rates
Anthropic announced today that the 1M token context window for Claude Opus 4.6 and Sonnet 4.6 is now generally available, with the long-context pricing premium completely eliminated.
China's Tech Giants Are Racing to Clone an Austrian Coder's AI Agent Tool
Four months ago, an Austrian developer named Peter Steinberger released OpenClaw, a free AI agent tool with a red lobster mascot. Today, it has triggered a full-blown corporate arms race in China that locals are calling "lobster fever."
xAI Loses 7 of 11 Co-Founders as Macrohard Coding Project Stalls
Only two of xAI's original eleven co-founders remain at the company.
How One CTO Uses Claude Code to Semi-Automate Every Code Review
Code reviews are still the bottleneck in most engineering teams, and they have been for decades. The tools change but the problem persists: someone has to read the diff, understand the context, run the tests, and decide if the code is good enough to ship. Grove CTO Daniel Olshansky thinks he has found a workflow that gets him most of the way there using Claude Code.
MCP Hits 97 Million SDK Downloads as Security Gaps Loom Over the Standard
Ninety-seven million monthly SDK downloads. Over 10,000 active server implementations. Fifteen months from specification to industry standard. Model Context Protocol, the open standard Anthropic published in November 2024 to let AI tools talk to external services, has crossed from "interesting experiment" into "thing everyone actually uses."
Google Closes $32B Wiz Acquisition, Largest VC-Backed Deal Ever
$32 billion. That's the final price tag on Google's acquisition of Wiz, now officially the largest venture-backed acquisition in history.
10,000 Authors Publish Blank Book at London Book Fair to Protest AI Training
"Don't Steal This Book" is 88 pages of author names followed by nothing. That blankness is the point.
Claude Code's Usage Limits Still Confuse Paying Subscribers
What do you actually get when you pay for Claude Code? That question keeps coming up, and Anthropic has not made the answer easy to find.
UberSKILLS: Open-Source Workbench for Authoring AI Coding Agent Skills
If you use AI coding agents like Claude Code, Cursor, or GitHub Copilot, you've probably run into SKILL.md files - markdown documents that teach these agents how to perform specific tasks within your codebase. Until now, creating them meant hand-writing YAML frontmatter and markdown with no preview, no validation, and no way to test before deploying.
OpenRouter Lists Hunter Alpha: 1 Trillion Parameters, 1M Token Context
One trillion parameters and a one million token context window (roughly 2,500 books worth of text). Those are the headline numbers for Hunter Alpha, a new AI model now listed on OpenRouter, the model routing service that lets developers access dozens of AI models through a single API.
Open-Source Library Adds Rule-Based Security Guards to AI Agent Frameworks
Most teams building AI agents today handle safety the same way: take a screenshot after the agent acts, send it to GPT-4, and ask "did this work and is it safe?" That approach is slow, expensive, and only as reliable as the model doing the judging.
Claude's Usage Limits Remain a Top Friction Point for Power Users
If you have spent any time building something ambitious with Claude, you have probably hit the wall. A long conversation goes dead. A complex prompt gets cut short. The model tells you it cannot process your full request. Claude's usage limits - both the context window (how much text the model can hold in a single conversation, roughly 150-200 pages for the largest window) and rate limits (how many messages you can send per hour) - remain one of the most common complaints from people trying to use it for real work.
How One Developer Got Claude to Work With 3D Geometry
Claude can write a React component or debug an API endpoint without breaking a sweat. Ask it to position a 3D object in space, and it falls apart. That gap between Claude's strong language skills and its weak spatial reasoning is something developer Dave Snider ran into head-first while building projects like Table Slayer and Counter Slayer - and his workaround is clever enough to be worth stealing.
Anthropic Puts $20M Behind Pro-Regulation Political Group for 2026 Midterms
Anthropic is spending $20 million to put AI regulation on the ballot. The company announced a donation to Public First Action, a new bipartisan 501(c)(4) organization that plans to back 30 to 50 candidates in state and federal races during the 2026 midterm elections.
Freelance Programmers Are Losing Clients to AI Coding Agents
"I had a customer meeting today, and the customer announced that he did not need me. He did everything he wanted in 2 days."
Meta Unveils Four Custom AI Chips, Plans Full Deployment by End of 2027
Four chip generations in two years. That's the pace Meta just committed to with its custom MTIA silicon program, and it signals a real shift in how Big Tech thinks about AI infrastructure costs.
Grammarly Pulls 'Expert Review' After Writers Find It Sold AI Impersonations of Them
Copy editor Benjamin Dreyer pasted several paragraphs of lorem ipsum into Grammarly's Expert Review tool. The AI returned writing advice attributed to Stephen King. Tips on improving placeholder text that isn't even English.
systemd Rejects Call to Ban AI Code, Maintainers Say Claude Finds Real CVEs
A GitHub issue demanding that the systemd project ban all AI-generated code contributions was closed today as "not planned," capping a heated debate that cuts to the heart of how open-source infrastructure projects are grappling with AI tools.
Agent.json Proposes a Robots.txt-Style Standard for AI Agents
Right now, if an AI agent needs to book you an appointment or request a refund on your behalf, it has to scrape a website, fill out a contact form designed for humans, or hope an API exists. There's no standard protocol for agent-to-website communication the way robots.txt handles crawler permissions.
Musk Recruits Senior Cursor Engineers as xAI Co-Founders Keep Leaving
Only 2 of xAI's 11 original co-founders remain at the company. That's the backdrop for Elon Musk's latest hiring push, which just pulled two senior engineers from Cursor, the AI coding tool that hit $2 billion in annualized revenue.
Claude's Rate Limits Have a Sync Problem That Locks Out Paying Users
You just got a fresh weekly quota on Claude. Zero percent used. But you still can't send a message because your session limit is maxed out from the previous burst - and it won't reset for another four hours.
Spine Swarm Puts Multiple AI Agents on a Visual Canvas for Business Tasks
What if instead of chatting with one AI in a text box, you could watch a team of agents work together on a whiteboard?
DashClaw Adds a Human Checkpoint Between AI Agents and Their Tools
What happens when your AI agent decides to delete a production database at 2 AM? Right now, most agent frameworks just... let it.
Apple's Updated Developer Agreement Sets New Rules for On-Device AI
Apple just updated its Developer Agreement with specific requirements for apps that use its on-device AI models, signaling how seriously the company is treating AI governance on its platforms.
Tendem Launches Hybrid AI-Plus-Human Task Outsourcing Platform
A new platform called Tendem is pitching a hybrid approach to task outsourcing: let AI handle the bulk of the work, then bring in human agents for the parts that require judgment.
VibeTrade: Open-Source Trading Agent Lets Claude Manage Your Broker Account
What happens when you give an AI model access to your brokerage account and tell it to trade? VibeTrade, a new open-source project, is built around that exact premise.
Building AI Agents That Know Who You Are, Not Just What You Want
Most AI setups remember your preferences. You like bullet points, you prefer concise answers, you work in marketing. But there is a meaningful difference between an agent that knows your preferences and one that knows who you are.
CanIRun.ai Checks Whether Your PC Can Handle Local AI Models
What GPU do you actually need to run Llama 3 or Mistral on your own machine? A new tool called CanIRun.ai tries to answer that question before you waste an afternoon on failed installations.
Three Studies, One Conclusion: LLM Agents Keep Failing Security Tests
Twenty AI researchers spent two weeks trying to break autonomous LLM agents in a live lab environment. The agents had email accounts, Discord access, file systems, and shell execution. The results, published in a paper titled "Agents of Chaos" by 38 researchers from Northeastern, Harvard, UBC, and CMU, document 10 distinct security vulnerabilities and 11 failure cases.
Google Deindexed VideoGamer.com After Its Pivot to AI-Generated Articles
A gaming publication just became the most visible casualty of Google's crackdown on AI-generated content. VideoGamer.com, a long-running gaming news site, has been completely removed from Google's search index after pivoting its editorial strategy to AI-written articles, according to a report from The Gamer.
TokenWatch Monitors AI API Spending Across OpenAI, Anthropic, and Google
If you've ever been surprised by an AI API bill at the end of the month, TokenWatch is built for exactly that problem. The new tool connects to your OpenAI, Anthropic, and Google Gemini API keys and provides a single dashboard showing real-time spend across all three.
ccStatusLine Tracks Your Claude Code Context Window Usage in Real Time
How much of your Claude Code context window have you burned through? Unless you've been counting tokens manually, you probably have no idea - and that matters more than you think.
AgentCard Gives AI Agents Prepaid Visa Cards with Built-In Spending Limits
The moment you let an AI agent buy things on your behalf, you face an obvious problem: how do you give it payment access without handing over your actual credit card? AgentCard's answer is purpose-built prepaid Visa cards that agents can create, fund, and spend through MCP (Model Context Protocol, the standard that lets AI tools like Claude interact with external services).
A Developer's Claude Code Setup: Custom Agents, Context Tracking, 40+ Skills
Freek Van der Herten, a prolific PHP developer at Spatie, has published his full Claude Code configuration - and it's a useful blueprint for anyone trying to get more out of the tool.
ChatGPT's Newer Models Have a Cliffhanger Problem
Have you noticed ChatGPT leaving you on a cliffhanger lately? You ask a direct question, get 80% of an answer, and then the model hits you with something like "Would you like me to explore this further?" or "There's another important dimension to consider here."
Context Engineering Is Replacing Prompt Engineering as the Real AI Bottleneck
Around 95% of generative AI pilots at large companies fail to deliver measurable return, according to MIT research. The reason isn't that the models are bad or the prompts are wrong. It's that the models can't see the right information at the right time.
Qwen 3.5-35B Runs Frontier-Class Code on a Single Consumer GPU
A year ago, running a model that could compete with Claude Sonnet or GPT-5 mini required API access and a monthly bill. Alibaba's Qwen 3.5-35B-A3B, released in February 2026 under the Apache 2.0 license, changes that math significantly.
Privacy Mask Redacts Secrets from Screenshots Before AI Agents See Them
Anyone who has shared a screenshot with an AI coding assistant has probably had that moment of panic: did I just send my API key to a cloud server?
Ads Are Coming to AI Chatbots. Here Is What That Actually Looks Like
$1.14 trillion. That is the projected size of the global advertising market in 2025, growing at 8.8% annually. AI chatbot companies want a piece of it, and they have started making moves.
Palantir Demos Reveal Pentagon Plans for AI Chatbots in Military Operations
Pentagon records and software demonstrations have surfaced showing how defense contractor Palantir is pitching AI chatbots as tools for military intelligence analysis and operational planning. Among the models featured in the demos: Anthropic's Claude.
OpenAI Offers Free ChatGPT Plus to Users Who Try to Cancel
Try to cancel your ChatGPT Plus subscription and OpenAI might just give it to you for free.
Developer's AI Agent Deletes 25,000 Documents From the Wrong Database
25,000 documents. Gone in seconds. Not from a cyberattack or a server failure, but from an AI agent that was told to clean up test data and picked the wrong database.
AI Coding Tools Can't Prove Your Code Is Correct - And That Gap Matters
Every AI coding assistant on the market today shares the same blind spot: they can write code that passes your tests, but they cannot prove that code is correct.
Lenovo's $4,099 ThinkStation PGX Puts 128GB of AI Compute on Your Desk
What does $4,099 buy you in local AI hardware right now? A box roughly the size of a hardcover book that can run 70-billion-parameter models entirely offline. Lenovo's ThinkStation PGX is the corporate-packaged version of Nvidia's GB10 Grace Blackwell chip, and it fills a gap that no consumer GPU currently can.
Microsoft Copilot Has a Familiar Problem: Nobody Picks It on Purpose
Microsoft has spent the last two years embedding Copilot into everything: Windows 11, Edge, Office 365, Bing, the taskbar, the Start menu. There is a dedicated Copilot key on new laptops. By pure distribution, Copilot should be the most-used AI assistant on the planet.
One Chat App, Four Frameworks: Pydantic AI Needs Half the Code of CrewAI
How much does your choice of AI framework actually cost you in code complexity? A new comparison from AI consultancy Vstorm puts a number on it: building the exact same chat application across Pydantic AI, LangChain, LangGraph, and CrewAI produced implementations ranging from 160 to 420 lines of code.
Tenstorrent's $9,999 QuietBox 2: 480 RISC-V Cores for Local AI Inference
476 tokens per second on Llama 3.1 70B. That is the headline number from Tenstorrent's QuietBox 2, a liquid-cooled desktop workstation that runs entirely on RISC-V processors and plugs into a standard 120-volt wall outlet. No rack. No server room. No specialized power.
Rick Beato Tells 5 Million Subscribers to Ditch Cloud AI for Local Models
A guitarist with five million YouTube subscribers just became one of the most visible mainstream voices arguing that local AI models will replace cloud-based ones. Rick Beato - best known for music theory breakdowns and industry commentary - published a video drawing a direct line between how the music industry failed and how commercial AI companies are headed for the same cliff.
Augment Code Replaces Coding Ability as Top Hiring Signal for Engineers
Writing code is no longer the top thing Augment Code looks for when hiring engineers. The AI coding tool company published a detailed framework this week that replaces traditional coding ability with six new evaluation dimensions, and the shift says a lot about where software engineering hiring is headed.
Solo Founder Hits $62k MRR in 90 Days Using AI Code Editors
$62,000 in monthly recurring revenue. Ninety days. One developer. No team, no co-founder, no venture backing at launch. Cameron Trew, a 26-year-old software engineer, quit his corporate job, moved back in with his parents, and built Kleo, a LinkedIn content tool, almost entirely with AI coding assistants.
LightPanda: A Non-Chromium Headless Browser Built for AI Agents
Most AI agents that browse the web today are dragging around the full weight of Chromium - a browser built for humans watching YouTube, not bots parsing DOM trees. LightPanda takes a different approach: it's a headless browser (a browser with no visible window, designed purely for automation) built from scratch in Zig, a low-level systems language, with nothing borrowed from Chrome, WebKit, or any existing browser engine.
Educators Say AI Is Hollowing Out Their Jobs From Both Ends
A community college tutor who spent years building trust with vulnerable students now spends her time helping them use tools that do the thinking for them. A programming instructor says almost every faculty meeting is about training teachers to use AI "rather than helping us teach students how to think and learn." An IT professional at Texas A&M was pushed out for refusing a Copilot mandate.
AI Coding Tools Write Great Functions but Can't Architect Your App
Hand an AI coding assistant a well-scoped task and it will deliver working code faster than any human. Ask it to build authentication, then settings, then an API layer across three separate prompts, and you get three internally coherent modules with completely inconsistent session handling between them.
Four AI Labs Independently Built the Same Agent Architecture
Cursor, Anthropic, OpenAI, and Google DeepMind built their AI coding systems independently. They ended up with nearly identical architectures. That convergence tells us more about where software engineering is headed than any single product announcement.
Meta Delays 'Avocado' AI Model Again, Now Targeting May at the Earliest
A $14.3 billion bet on one hire, 70-hour workweeks, and a string of delays. That is the current state of Meta's flagship AI effort.
Adobe CEO Shantanu Narayen to Step Down After 18 Years as Shares Drop 7%
Shantanu Narayen, the CEO who turned Adobe from a boxed-software company into a $25 billion cloud giant, announced on March 12 that he will step down once the board names a successor.