AI News
AI news that matters. Updated daily.
No stories match your filters.
Miasma Traps AI Scrapers in Infinite Loops of Poisoned Content
Most website owners fighting AI scrapers play defense: block user agents, rate-limit requests, throw up CAPTCHAs. Miasma takes the opposite approach. Instead of keeping scrapers out, it invites them in and wastes their time with an infinite maze of garbage content.
Google Releases Scion, a Testbed for Running Multiple AI Agents in Parallel
Running one AI coding agent is straightforward. Running five of them at once, each with its own credentials and workspace, without them stepping on each other? That's the problem Google Cloud Platform built Scion to solve.
Developer AI Spending Data: Anthropic Costs More but OpenAI Gets More Usage
$1.87 million. That's how much 572 developers collectively spent on AI coding tools, according to usage data tracked by tokscale, an anonymous logging tool that reads local usage data from AI coding assistants without capturing API keys or actual code.
DaVinci-MagiHuman: Open-Source 15B Model Generates Talking Head Video in 2 Seconds
A new open-source model called daVinci-MagiHuman can generate a 5-second lip-synced talking head video in 2 seconds on a single H100 GPU. The model, jointly developed by Sand.ai and SII-GAIR Lab, ships under Apache 2.0 - meaning anyone can download, modify, and use it commercially with zero restrictions.
ai-setup Auto-Generates CLAUDE.md and Cursor Rules From Your Codebase
Anyone using AI coding assistants knows the ritual: before the AI can help you, you have to explain your entire project to it. Write a CLAUDE.md. Set up .cursorrules. Configure Windsurf rules. It is busywork that defeats the purpose of having an AI assistant in the first place.
ChatGPT Pro Users Report Account Bans With No Warning or Explanation
ChatGPT Pro subscribers are reporting sudden account bans with no prior warning, no explanation, and no clear path to resolution.
Non-Coders Are Building Real Web Apps With AI - And Can't Stop
A year ago, "I don't know how to code" meant you couldn't build software. Now it means you haven't tried yet.
The CLI Is the Reason Claude Code Dominates AI Coding Tools
Every other AI coding tool is trying to build the perfect GUI. Claude Code shipped a terminal prompt and started winning.
The Case Against AI in Production Isn't Stupid, But It Is Incomplete
"Why isn't using AI in production considered stupid?" It's a question that keeps surfacing in developer communities, and honestly, it deserves a better answer than the industry usually gives.
Shoofly Adds a Security Layer Before AI Coding Agents Execute Commands
36% of community-shared skills for AI coding agents contain security flaws, according to a Snyk audit. Trend Micro has documented actual malware being distributed through skill marketplaces. And Anthropic's own safety page states plainly that their built-in filters "are not a security boundary."
A Platform Engineer's Case Against Agentic SRE and AI for Infrastructure
A detailed field report from a platform engineer makes a sharp argument: AI agents are nowhere close to handling real infrastructure work, and the industry is kidding itself by pretending otherwise.
How to Spot an AI Slop Website (And Avoid Building One Yourself)
Building a full website with Claude or ChatGPT takes about 20 minutes now. Making one that doesn't scream "AI built this" takes considerably longer, and most people skip that second part.
Did Meta's $50B VR Bet Cost Zuckerberg the AI Race?
Between 2020 and 2025, Meta spent over $50 billion on Reality Labs, its virtual and augmented reality division. During that same period, OpenAI launched GPT-4, Google shipped Gemini, and Anthropic built Claude into a serious competitor. The argument that Zuckerberg's metaverse fixation cost Meta its shot at AI leadership is gaining traction among prominent tech investors.
Why Spawning Multiple AI Agents from One Model Won't Give You Diverse Thinking
"Hire AI employees with different viewpoints on the same task." It sounds like a reasonable prompt. Tell Claude or GPT to act as an orchestrator, spawn sub-agents with different roles, and get genuine diversity of thought.
Eval-Driven Development Brings TDD Discipline to AI Prompt Engineering
Most teams building AI agents today are still doing prompt engineering by vibes. Change a word, eyeball the output, decide it "looks better," ship it. It's the equivalent of writing software without tests and hoping nothing breaks in production.
GitAuto Claims Its Test Generator Catches What Plain Claude Misses
Most developers who've used Claude for writing unit tests know the pattern: you get clean, well-structured tests that cover the obvious paths and call it a day. GitAuto, a startup focused on automated test generation for GitHub repos, published a comparison showing where that approach falls short.
Phantom Secrets: Open-Source Tool Keeps API Keys Hidden from AI Agents
A new open-source project called Phantom Secrets addresses a growing concern in AI agent workflows: your API keys shouldn't be visible to the model itself.
Wikipedia's English Edition Votes 40-2 to Ban AI-Generated Articles
Wikipedia's English-language edition has officially banned AI-generated content. Volunteer editors voted 40 to 2 in favor of a new policy that prohibits using large language models (LLMs) to create or rewrite articles on the platform.
Google AI Mode Caught Serving Conspiracy Theories in Search Results
Google's AI Mode - the company's AI-generated answer feature built into Search - returned conspiracy theories instead of factual information when a user asked about the death of Food Network chef Anne Burrell.
SourceBridge: Open-Source Tool That Maps and Explains AI-Generated Codebases
What happens when your team has been using AI coding assistants for months and nobody fully understands what got built? SourceBridge is a new open-source project that tries to solve exactly that problem - generating documentation, architecture diagrams, and guided code tours from any codebase automatically.
No AI Can Reliably Calculate Your Taxes - Best Model Scores Just 32%
23 to 32%. That's the range of fully correct federal tax returns produced by today's best AI models when tested against real IRS forms. If your accountant had that accuracy rate, you'd fire them immediately.
56% of Power Users Funnel All Messages Into One AI Thread
56% of the most active users on one AI companion platform put over 70% of their messages into a single conversation thread. Not multiple topics, not different characters - one ongoing relationship, deepened over months.
Anthropic Publishes Multi-Agent Blueprint for Longer Claude Coding Sessions
Anthropic just published a detailed guide on building what it calls a "harness" around Claude for extended coding work, and it directly addresses two problems that anyone using AI for serious development has run into.
Could $9 Trillion in AI Data Center Spending Become Tech's Biggest Bust?
$9 trillion. That's the figure now being attached to global AI data center investment commitments, and the Financial Times is asking the question that many in the industry have been avoiding: what if demand never catches up?
Anthropic Says Claude Has Found Over 500 Zero-Day Vulnerabilities in Production Software
A year ago, AI-assisted security research was mostly a novelty. Now Anthropic's Frontier Red Team says Claude has found and validated more than 500 high-severity zero-day vulnerabilities (previously unknown security flaws) in production open-source software, with no specialized tooling, custom scaffolding, or hand-crafted prompts.
One Photo Is All It Takes: AI Tools Chain Together Into a Surveillance Kit
What does it take to find where someone lives, starting from a single photo? About ten minutes and a handful of free tools.
ChatGPT's Simplified Sidebar Sparks Fresh User Backlash
OpenAI recently rolled out a simplified sidebar on ChatGPT's mobile apps, and the reaction from daily users has been predictably hostile. Features like Images, Codex, Pulse, and Apps now sit in a horizontal bar above your chats and projects instead of living in the sidebar menu. The idea is to free up space for your conversation history. In practice, plenty of users feel like they got a downgrade.
Local LLM Inference Runs Nearly 2x Faster on Linux Than Windows
If you're running local AI models on Windows and wondering why they feel sluggish, the operating system itself might be the bottleneck.
Developer Translates Two Full Codebases With Claude, Gets Byte-Perfect Output
Zero divergences across 321,331 data points. That's the result when developer Daniel Janus pointed Claude Opus 4.6 at two non-trivial codebases and asked it to translate them into entirely different programming languages.
AI Coding Agents Have a Cost Control Problem Nobody Solved Yet
A developer's AI coding agent hits an error, retries the same failing approach four times, and burns through $12 of API calls in three minutes. The developer doesn't find out until the end-of-month bill arrives, because every tool on the market shows the same thing: total tokens, total cost, maybe a per-model breakdown. Nothing about which agent, which task, or which runaway retry loop ate the budget.