AI News Daily

Spotify and Universal Music Group have struck a deal that lets Premium subscribers create AI-generated covers and remixes of songs, with participating artists receiving a share of the resulting revenue.

This is the first major deal between a streaming platform and a major label that explicitly permits fan-created AI music — with a payment structure attached. Instead of blocking AI covers outright, Universal is licensing the capability and collecting royalties when fans use it.

The deal is significant because it creates a model: AI-generated music built on real artists' work can coexist with those artists getting paid. Whether fans actually use the feature at scale remains to be seen, but the legal framework is now in place.

Anthropic launched an official directory of Claude Code plugins — tools and extensions that give Claude new capabilities when working in codebases. The repository is managed directly by Anthropic and sets a quality bar for what plugins should look like.

Before this, the Claude Code ecosystem was growing fast but informally — developers shared plugins across GitHub and Discord without a central place to find the best ones. This directory changes that. It gives developers a vetted starting point and gives Anthropic a way to signal which integrations it supports.

The repository has attracted strong interest from developers building with Claude Code who want a reliable set of first-party tools to start from.

Anthropic has told its investors that it is on track to more than double revenue in the second quarter of 2026, reaching around $10.9 billion. This would mark the first time the company has turned a profit since it was founded.

The number is striking. Anthropic has been spending heavily on model training and compute, and for years operated at a large loss. A jump to nearly $11 billion in quarterly revenue — and profit — is a different kind of story entirely.

The disclosure came as part of investor communications and was reported earlier this week. Anthropic has not issued an official public announcement, but the signal is clear: Claude has become a genuine commercial product at serious scale.

President Trump pulled back from signing an executive order that would have required government security reviews of AI models before they could be released. The order had been drafted and was ready to sign, but Trump delayed it, saying the language could have acted as a roadblock to American AI development.

The original order would have put new checks on the most powerful AI systems — requiring labs to submit models for review before deploying them publicly. That kind of requirement has been a flashpoint in Washington, with AI companies pushing hard against anything that slows their release timelines.

The delay is a win for AI labs in the short term. It also signals that the administration is more focused on keeping the US ahead in the AI race than on building safety guardrails around frontier models.

Nvidia CEO Jensen Huang says his company has identified a new $200 billion market: CPUs designed specifically for AI agents. This is distinct from the GPU business that has driven Nvidia's recent growth. Agents need different hardware — processors built to handle the continuous, parallel workload of autonomous software running tasks in the background.

Huang made the comments as Nvidia reported another record quarter, with $43 billion in holdings across AI startups also disclosed in the same financial update.

The framing matters. Huang is not talking about building better GPUs. He is talking about a whole new chip category for a whole new computing era — one where AI agents run constantly, not just when a human types a prompt.

CodeGraph is a tool that builds a pre-indexed knowledge graph of your entire codebase, then serves it to AI coding agents like Claude Code, Cursor, and Codex. Instead of forcing the AI to explore your project file by file on every session, CodeGraph builds the map once and keeps it updated.

The result is fewer tokens used per task and fewer tool calls made by the agent. For large codebases, this can mean the difference between an AI that fumbles around and one that knows exactly where things are.

The project runs entirely on your own machine — no cloud, no data leaving your system. It has quickly become one of the most-starred new developer tools in the community, with builders reporting measurable speed improvements on complex coding tasks.

This project distills Andrej Karpathy's observations about common AI coding mistakes into a single CLAUDE.md file you drop into any project. The file gives Claude Code a set of guardrails and preferences drawn directly from Karpathy's public writing about where AI coding tools tend to go wrong.

Karpathy is one of the most respected voices in AI — a former director at OpenAI and the person who popularized many current ideas about how to work with language models. His observations about coding pitfalls are specific and practical, not generic advice.

The project has attracted enormous attention in the developer community. For teams already using Claude Code, dropping this file into a project is a low-effort way to apply some of the clearest thinking available on how to get better results.

Matt Pocock, one of the most followed TypeScript educators in the developer community, published his personal Claude skills collection — the exact set of skills he uses in his own .claude directory. Unlike generic prompt libraries, these are opinionated tools built around real engineering workflows.

The collection covers planning, automation, and agent-ready API patterns, with each skill reflecting how a working engineer actually structures AI-assisted development rather than how demos make it look.

Pocock has built a large following by making complex TypeScript concepts accessible. This release brings the same philosophy to AI tooling: real patterns from real usage, shared directly.