AI News
What actually shipped this week, checked against the primary source rather than the headline about it.
Sep 26, 2026
The AI Model Price War Arrived in a Two-Hour Window
Claude Opus 5.5 and GPT-6 Sol and Luna both launched September 22 with steep cuts. Here's what the AI model price war actually costs and what it means for your bill.
Sep 26, 2026
Amazon Blocks AI Shopping Agents: Inside the Meta Muse Standoff
Amazon blocked AI shopping agents from Meta's Muse over the weekend, the first real test of who gets to check out on your behalf and on whose terms.
Sep 26, 2026
ChatGPT's Ad Pixel Tracks What You Do on Other Sites, and Advertisers Can't See It
A researcher's writeup shows ChatGPT's ad pixel tracking sets a cross-site cookie that links your browsing to your ChatGPT account, invisible to the advertiser who installed it.
Sep 26, 2026
Claude Code Cloud Sessions Are Out of Preview, With a Credit Deadline Attached
Claude Code cloud sessions left research preview on September 23. Here's how claude --cloud actually works, and the October 7 deadline on Anthropic's trial credit.
Sep 26, 2026
Claude Discovers a Novel Enzyme System: What 950 Agents Actually Did
Claude discovers a novel enzyme system with CRISPR-like repeats after 950 agents read 200,000 enzymes in 21 hours. What they did, what humans did, what's unknown.
Sep 26, 2026
Qwen-Image 2.1's License Is the Real Story, Not Its Feature List
Qwen-Image 2.1's license restricts the model to non-commercial use, a reversal from Alibaba's earlier Apache 2.0 image releases, and its version number is misleading.
Sep 20, 2026
Claude Code Projects: One Conversation Now Coordinates Parallel AI Threads
Claude Code Projects turns one conversation into a coordinator that splits work across parallel cloud threads. Here is how it actually works and what it costs.
Sep 20, 2026
Microsoft's Own Memo Called AI Training Data Scraping 'The Largest Theft in History'
Unsealed court filings show a Microsoft director privately calling AI scraping the largest theft of labor in history. Here is what the documents reveal.
Sep 20, 2026
Salesforce in Claude: What Beta Access Actually Gets You
Salesforce in Claude is now in open beta with 37 sales skills built in. Here's what the plugin connects, how approval works, and who's already using it.
Sep 14, 2026
What Google Merchant Center's Leaked AI Prompts Reveal About Building AI Features
Google Merchant Center briefly leaked its own AI prompts, showing the exact structure, guardrails, and quality check behind a real production AI feature.
Sep 14, 2026
OpenAI's Navier-Stokes Proof Comes With a Credit Fight Attached
OpenAI's Navier-Stokes proof used 10,000 agents to crack a $1M math problem in 88 hours, then a credit dispute with two mathematicians overshadowed it.
Sep 5, 2026
Claude's Background Computer Use Is Live, With a Commerce Agent Blueprint Alongside It
Claude background computer use now runs on Mac and Windows while you work elsewhere. Here's what it does, its safety limits, and what shipped with it.
Sep 5, 2026
Claude Fable 5.1 and Mythos 5.1: Same Model, Different Rules
Anthropic shipped Claude Fable 5.1 and Mythos 5.1 together on September 1. Same underlying model, one gated behind a verification program. Here's what changed.
Aug 29, 2026
OpenAI's Jalapeño Chip Benchmark Numbers Beat Nvidia's Blackwell
OpenAI's first Jalapeño chip benchmark results beat Nvidia's Blackwell on inference efficiency, but the company's own data says Blackwell isn't the fair comparison.
Aug 23, 2026
ChatGPT's Reddit Citation Drop Isn't as Settled as It Looks
ChatGPT's Reddit citations reportedly dropped 86% this month, but two tracking vendors disagree 4x on the baseline and the timing doesn't match the blamed cause.
Aug 20, 2026
How OpenAI Plans to Catch Misuse Without Breaking Zero Data Retention
OpenAI previews Private Safety Processing, a way to flag misuse patterns across sessions under zero data retention without staff seeing your prompts or responses.
Aug 19, 2026
Claude Code's New /design Skill Puts Claude Design's Artboards in Your Terminal
Anthropic's new Claude Code design skill brings Claude Design's artboard workflow into the CLI, so you pick a UI option and have Claude build it directly.
Aug 19, 2026
DeepSeek Harness Ships Free, Days Before a Steep API Price Hike
DeepSeek open-sourced Harness, a free Claude Code-style agent framework, then confirmed API price increases of up to 1,100 percent on V4-Pro two days later.
Aug 19, 2026
Gemini 3.7 Flash Launches at Half Price, With a Hard Expiration Date
Gemini 3.7 Flash pricing runs at half price only through December 2026. What actually improved, the real benchmark numbers, and what the deadline means.
Aug 17, 2026
Stripe Is Reportedly Buying OpenRouter for Over $7 Billion
Stripe is reportedly paying over $7 billion in the Stripe OpenRouter acquisition, buying the layer that decides which AI model handles each request.
Aug 13, 2026
Claude in Chrome Now Carries Your Cowork Session Across Desktop, Web, and Mobile
Claude in Chrome Cowork sessions now sync across desktop, web, and mobile, and the browser extension can click, type, and fill forms using your own logins.
Aug 12, 2026
Claude's New Text Watermark Changes What 'AI-Written' Means
Anthropic now embeds an invisible claude text watermark in new model output under the EU AI Act. Here is what it covers, what it proves, and what it doesn't.
Aug 11, 2026
Claude Code Auto Mode Becomes the Default on August 14: What Actually Changes
Claude Code auto mode becomes the default for Pro, Max, and Team plans on August 14. Here is how the safety classifier works and what changes for you.
Aug 10, 2026
Oracle Wrote Two AI-Generated Code Policies. They Contradict Each Other.
Oracle's OpenJDK bans AI-generated code outright while its own GraalVM project allows it. The split is a working template for writing your team's AI code policy.
Aug 7, 2026
This Week in AI: August 7, 2026
The AI updates worth your attention this week, and what each one means for your system.
Aug 5, 2026
Google Cloud's 2027 Revenue Will Depend on Two Money-Losing AI Companies
UBS estimates say OpenAI and Anthropic will drive 48% of Google Cloud's 2027 revenue, worth $124 billion, though both companies are losing billions yearly.
Aug 5, 2026
Qwen3.8-Max Beat 458 of 526 Human Teams in a Live Data Contest
Alibaba's Qwen3.8-Max autonomously beat 458 of 526 human teams in a live data contest, the third open-weight model in ten days to claim a top benchmark spot.
Aug 3, 2026
DeepSeek's V4-Flash Benchmarks Jumped Without a Bigger Model
DeepSeek V4-Flash benchmarks rose sharply after a retrain, not a size increase, and the company says it now exceeds its own flagship Pro model.
Jul 30, 2026
AI Agents Followed Company Handbooks Only 36% of the Time
The HANDBOOK.md benchmark tested AI agents against realistic company policy documents up to 124 pages long. The best model followed them correctly only 36% of the time.
Jun 21, 2026
This Week in AI: June 21, 2026
The AI updates worth your attention this week, and what each one means for your system.
Jun 13, 2026
This Week in AI: June 13, 2026
The AI updates worth your attention this week, and what each one means for your system.
Jun 6, 2026
This Week in AI: June 6, 2026
The AI updates worth your attention this week, and what each one means for your system.