AI News
AI news that matters. Updated daily.
No stories match your filters.
DeepSeek Raises $10.29 Billion, Founder Pledges Open-Source AI Development
$10.29 billion. That's the size of DeepSeek's latest financing round, according to Bloomberg reporting, making it one of the largest single AI funding rounds globally. DeepSeek founder Liang Wenfeng declared an AGI goal - AGI meaning AI systems capable of human-level reasoning across general tasks, not just narrow benchmarks - alongside a stated commitment to continue building open-source models rather than pivoting toward short-term revenue.
AI Reconstructed Dead Pilots' Voices from Public NTSB Spectrograms
The NTSB publishes accident investigation dockets online - thousands of pages of evidence, transcripts, and technical data that researchers, journalists, and families use to understand what went wrong. Those dockets sometimes include spectrograms: visual charts that map sound frequencies over time, released as images rather than audio files. The agency has treated this as a reasonable middle ground for years. You can see the data without hearing the voices of the people who died.
Claude Opus 4.7 Is Pushing Back on Tedious Work and Recommending Better Tools
What happens when an AI model decides your request isn't worth doing manually?
NVIDIA Drops Gaming as a Standalone Revenue Category in Financial Reports
For most of its history, NVIDIA's gaming business was the company's defining product line - the reason enthusiasts built custom rigs and why the GeForce brand became synonymous with performance. That era has effectively ended, at least in terms of how NVIDIA presents its finances.
Anthropic's Project Glasswing Claims 10,000+ Critical Security Bugs Found in 30 Days
10,000 critical vulnerabilities in one month. Anthropic is claiming that figure for Project Glasswing, its security research initiative that deploys Claude to scan codebases and infrastructure for exploitable flaws.
NTSB Pulls Crash Docket After AI Users Clone Dead Pilots' Voices
Public crash investigation records exist because aviation safety depends on transparency. But the same cockpit voice recordings that help investigators reconstruct what went wrong in a fatal accident now provide enough raw audio for AI tools to rebuild the voices of pilots who didn't survive.
Uninstall GSD AI Tool Now: Creator Ran a Crypto Rug Pull, Still Has NPM Access
The "Get Shit Done" (GSD) AI tool has been abandoned by its creator, who launched a $GSD cryptocurrency token tied to the project, collected the proceeds, and disappeared. If you have the original NPM packages installed, remove them now. The original developer still has publish rights to those packages, which means they can push a malicious update that runs silently on your machine the next time you install or upgrade.
Microsoft Drops Claude Code Over API Cost Overruns
Microsoft pulled the plug on its internal use of Claudee Code(/tools/claude-code/) after the AI coding assistant ran over budget, according to reports. The decision is worth paying attention to given who made it.
Grok Barely Shows Up in Federal AI Records, Reuters Report Finds
Federal agencies logged their AI tool usage last year, and Grok barely showed up. That's the finding of a new Reuters report, which combed through government records and found xAI's chatbot - Elon Musk's self-described "truth-seeking" AI - almost entirely absent from official federal use. The Verge covered the Reuters findings and paired them with a blunt conclusion: the chatbot is neither especially good nor especially popular.
Claude as a Devil's Advocate: The Strategic Use Case Most People Miss
Most people open Claude and ask it to write something. The practitioners getting the most out of it are using it to argue back at them.
Google Search Breaks When You Type 'Disregard'
Type "disregard" into Google Search and the interface breaks. No normal results - just a failed experience that suggests the AI powering Google's search summaries is treating a common English word as a threat.
Typing 'Disregard' Into Google Search Breaks AI Overviews Into Chatbot Mode
What happens when you search for the word "disregard" in Google right now? Instead of the usual AI Overview - the auto-generated summary box that appears above search results - Google's AI switches into full chatbot mode, responding as if you're having a conversation rather than looking something up.
Google's Android XR Glasses Use Gemini for Live Translation and Navigation
Real-time translation overlaid on your actual field of view. Turn-by-turn navigation without pulling out your phone. Google's prototype Android XR glasses, powered by Gemini AI, can do both of those things. They're just not shipping yet.
Google AI Search Is Convenient Enough That You'll Use It Anyway
The fastest way to get someone to use a product they distrust is to make it just slightly faster than the alternative. That's Wired's diagnosis of where Google AI search is headed - and it's hard to argue with.
An AI-Written Story Apparently Won a Commonwealth Prize. Nobody Caught It.
The Commonwealth Short Story Prize is not a hobbyist contest. It draws entries from across the Caribbean, Canada, Africa, Asia, Pacific, and Europe, and regional winners get published in Granta, the British literary magazine that has run since 1889. This year, one of those winners appears to have been written by AI.
Spotify and UMG Launch AI Remix Tool as AI Covers Already Flood the Platform
The reggae version of "Smells Like Teen Spirit" nobody asked for has been on Spotify for months. So have the AI country renditions of The Weeknd, the Motown reimaginings of AC/DC, and hundreds of other algorithmically generated covers that clog up search results when you just want to find the original. Spotify, in partnership with Universal Music Group, just built a tool that makes generating them faster.
Claude Max vs ChatGPT Pro: One Accountant's $100/Month Verdict
$100 a month is what both Anthropic and OpenAI charge for their top consumer tiers. Claude Max and ChatGPT Pro cost exactly the same, which makes direct comparisons from working professionals unusually useful.
AI-Generated Stories Secretly Won 3 of 5 Fiction Awards
3 out of 5. That's how many fiction awards went to AI-generated stories in a competition where judges had no idea the entries weren't written by humans.
Anna's Archive Writes an Open Letter to AI Models - and Offers Them a Deal
What do you do when AI companies have almost certainly trained models on your data without asking? Anna's Archive has an answer: publish an llms.txt file and negotiate directly.
Sam Altman Offers Every Y Combinator Startup a Deal They Can't Ignore
Sam Altman, who ran Y Combinator before taking the CEO role at OpenAI, has made an offer to every startup in the YC portfolio. The framing - widely described as a "mic drop" moment - suggests the terms are significant enough to turn heads across the startup community.
The MCP Servers Actually Changing How People Use Claude
Six months ago, the Claude workflow was: find file, copy relevant section, paste into chat, hope it fits within the context window (the maximum amount of text Claude can process at once - roughly a 300-page book on most plans), get answer, go back to work. Repeat every session.
Token-Based AI Billing Is Breaking Enterprise Budgets. Microsoft's Anthropic Licenses Are One Casual
Annual budgets set in January don't survive contact with token-based AI billing. Microsoft reportedly canceled internal licenses for Anthropic's Claude after usage-based costs blew through full-year allocations within months - a pattern that's becoming an uncomfortable reality for enterprises moving from flat-rate seat licenses to pay-per-token models.
Claude API Had Elevated Error Rates Early May 22
Anthropic logged an elevated error rate on the Claude API (the service developers and businesses use to connect their apps to Claude) starting at 4:16 AM UTC on May 22, 2026. The incident was tracked on Anthropic's public status page.
Spec-Driven Development: A Claude Code Workflow That Clears Context Between Steps
What if the biggest productivity problem with AI coding agents isn't the model, but how you structure the conversation with it?
AI Smart Glasses Have an Infrastructure Problem the Hardware Can't Solve
Meta's Ray-Ban smart glasses sold over a million units in 2024. Google is reportedly building a new version of its smart glasses. Samsung is working on AI-integrated frames. The hardware momentum is real. But the argument that these devices are running well ahead of the infrastructure needed to support them deserves serious attention.
Samsung Chip Workers Average $340K Bonus as AI Hardware Profits Surge
$340,000. That's the average bonus Samsung's chip division workers are set to receive as the company's semiconductor business rebounds on AI demand, according to a Quartz report.
Gemini Gives a Candid Answer When Asked Why Google Search Has Gotten Worse
When a frustrated user asked Gemini - Google's own AI assistant - why Google Search has gotten so bad, the response was more candid than you'd expect from a product built by the same company.
OpenAI's New Political Chief Wants to Calm the AI Regulation Fight
The AI safety debate has grown loud enough that it's now a political liability for OpenAI. The company's response: hire the Democratic crisis manager who helped navigate the Clinton White House through the Lewinsky scandal.
OpenAI's Codex Named a Gartner Leader in Enterprise AI Coding Agents
Gartner placed OpenAI in the Leader quadrant of its 2026 Magic Quadrant for Enterprise AI Coding Agents, with Codex cited for innovation and enterprise-scale deployment capability.
Virgin Atlantic Hit Its Holiday App Deadline Using OpenAI's Codex
Fixed deadline. Holiday travel season. A mobile app that couldn't ship late. That's the scenario Virgin Atlantic's engineering team brought to OpenAI's Codex - and the case study they published makes for a useful real-world data point on where AI coding agents actually deliver.
Meta Built 4 Custom AI Chips in Two Years to Cut Inference Costs at Billion-User Scale
Four generations of custom AI silicon in two years. That's the pace Meta has set with its MTIA (Meta Training and Inference Accelerator) program, which the company detailed in a recent blog post. Most chip programs take three to five years between major generations. Meta is running at roughly double that speed.
Meta's Segment Anything Model Powers Alta Daily's Digital Wardrobe Feature
A fashion startup called Alta Daily is using Meta's open-source image-recognition technology to build a digital wardrobe catalog, and the implementation shows how AI research tools are finding their way into practical consumer products.