1. Anthropic Shipped Claude Code's Full Source in an NPM Update A developer on X spotted it first: a .js.map file sitting inside Claude Code's version 2.1.88 NPM package.

2. OpenAI Raises $122 Billion at $852 Billion Valuation as Product Ambitions Narrow OpenAI closed a $122 billion funding round on March 31, the largest private raise on record, setting its valuation at $852 billion.

3. Writers and Art Students Reach the Same Breaking Point Over AI A multilingual blogger confesses they can no longer write a thousand words without reaching for an AI tool.

In Brief

• Police Wrongly Arrested Tennessee Woman After AI Facial Recognition Match Police in North Dakota used AI facial recognition to identify Angela Lipps, a Tennessee resident, as a suspect. She was arrested despite having no connection to the state or the alleged crimes.

• Google Releases Veo 3.1 Lite, Its Cheapest Video Generation Model Veo 3.1 Lite is now in paid preview through the Gemini API and available for testing in Google AI Studio. Google positions it as its most cost-effective option for developers building video generation into apps.

• MIT Technology Review Calls for Replacing Human-vs-AI Benchmarks AI benchmarks have long measured whether models outperform humans on isolated tasks like chess, math, or coding. That framing fails as models move into open-ended, real-world work where no single-human baseline applies. The piece argues evaluation must shift to deployment-specific metrics.

• ChatGPT Now Runs on Apple CarPlay Dashboards Apple's iOS 26.4 update added support for voice-based conversational apps in CarPlay. OpenAI shipped a ChatGPT integration immediately, making it the first AI chatbot accessible from the car dashboard.

• Hugging Face Paper Proposes First Autonomous AI System for Clinical Research Medical AI Scientist generates clinical hypotheses, designs experiments, and drafts manuscripts. Existing autonomous research agents lack the medical domain grounding and specialized data pipelines this framework provides.

• Samsung Galaxy S26 Ships Open-Ended AI Photo Editing Samsung's default photo app now accepts natural-language requests to alter images. The feature extends beyond background tweaks into full scene manipulation, following Google's Pixel 9 AI editing tools.

• Study Catalogs Failure Modes When Multiple AI Agents Negotiate Together Multi-agent systems built on large generative models produce risks that no single agent would exhibit alone. The paper identifies emergent failures in collective planning, resource allocation, and negotiation at scale.

• MIT Technology Review: Domain-Specific AI Models Now Deliver Bigger Gains Than General Upgrades General-purpose LLM improvements have flattened to incremental jumps. Models fused with organization-specific data still produce large capability leaps in specialized tasks.

Read the full edition →