The AI Friday Brief
Good Wednesday, NOLA. June 3rd is consolidation day after yesterday's IPO shock. The Economist digs into whether the market can handle three mega-IPOs at once, while GitHub launched Copilot as a standalone app and Microsoft quietly shipped Scout, a new autonomous agent. Plus: AI is outperforming law professors, Google's getting regulatory heat on search, and there's a fascinating debate about whether RSS is making a comeback for AI agents.
Big Moves & Context
Can the stock market swallow Anthropic, SpaceX, and OpenAI?
Popular on HN. The Economist digs into the math: three mega-cap IPOs in one year is wild. Anthropic's filing yesterday is the headline, but this zooms out on whether we're in a bubble or if the valuations actually make sense. Worth reading before the roadshow chaos begins.
The Economist
Microsoft Scout: Autonomous AI agent built on OpenClaw
Microsoft's taking a harder swing at autonomous agents with Scout, built on their OpenClaw infrastructure. This is the move we flagged yesterday as coming — the big vendors aren't just wrapping existing models anymore, they're building agent-first products. HN discussion here.
Hacker News
AI outperforms law professors in Stanford Law study
Law professors just lost a benchmark. In a direct comparison, AI outperformed practicing attorneys on case analysis. It's the kind of study that makes real people in real industries take a hard look at their workflows. HN discussion.
Stanford Law School
Tools & Products
GitHub Copilot gets its own app (preview)
GitHub just launched Copilot as a standalone application, moving it beyond just VS Code and the browser. This is a big signal: they're betting people want a dedicated AI coding interface, not just a suggestion engine inside their editor. HN discussion.
GitHub
Paseo – Open-source AI coding agent interface
Show HN from the community: a beautiful, open-source UI for AI coding agents. If you're building agent workflows or want a nicer interface than the CLI, this is worth a look. Discussion here.
Hacker News
Building Search for AI Agents with Exa CEO Will Bryk
Great conversation on the a16z AI podcast. Exa's CEO talks about why traditional search is broken for agents and what search infrastructure actually needs to look like when the end-user is code, not a human. Worth a listen if you're building anything that needs to find information reliably.
a16z AI
Regulation & The Real World
Google must let publishers opt out of AI Search features, rules UK
The UK's CMA just handed publishers a win: they can now opt out of Google's AI Search features. This is the first real regulatory punch at AI integration, and it matters because it sets a precedent. If the UK can do this, others will follow.
The Verge
Trump signs downsized AI order after weeks of reversals
Discussed on HN. After weeks of back-and-forth, the White House released a much narrower AI policy order than initially planned. Less dramatic than the hype, but worth understanding what actually made it through.
Politico
Interesting Deep Dives & Reads
Now AI agents need what RSS does
Fascinating take: RSS is coming back, but not for humans — for AI agents. As agents get smarter, they need reliable feeds of structured information. This is a genuinely clever insight about plumbing that's going to matter a lot. HN thread here.
Hacker News
How we index images for RAG
Kapa's team wrote a solid technical guide on making image search work in retrieval-augmented generation (pulling relevant images into context for AI models). If you're building anything that needs visual understanding at scale, this is practical. Discussion.
Hacker News
AI has a water problem. Google thinks it has a fix.
As data centers scale up, water consumption is becoming a real crisis. Google's announcement on water-positive commitments matters not just for environmental reasons, but because it signals the industry can't ignore infrastructure costs anymore. Related: California's universities have wildly different AI policies.
The Verge
Also
- Adafruit receives demand letter from legal counsel on behalf of Flux.ai — Legal dispute emerging in the AI hardware space — worth watching.
- Agentic Motherfucking Website — Playful take on what 'agentic' actually means. Fun read.
- Bringing Up DeepSeek-V4-Flash on AMD MI300X — Practical guide to running large models on AMD hardware. Useful if you're exploring alternatives to NVIDIA.
- More than 6 out of 10 people turn to AI for psychological support — Behavioral shift: people are using AI as a mental health tool at scale. Interesting data from AXA's health report.
- Microsoft wants to 'make people addicted' to Scout — Internal docs reveal Microsoft's strategy for Scout adoption. The language is wild.
- U of T researchers demonstrate AI worm could target any online device — Security research showing AI as a vector for attacks. Serious but not doomsday.
- AI engineers aren't safe from being replaced by AI — Thoughtful essay on job displacement. Less panic, more realism.
- Different attitudes towards AI in California's university system — Long-read on how different UC schools are handling AI in classrooms. Surprising divergence.
- Walmart caps usage of AI tool for employees after high demand — Real-world adoption friction: even at scale, AI tool demand can overwhelm infrastructure.
- Nvidia introduces first PCs designed for AI agents — Hardware is shipping now for agent workflows. The PC market is shifting.
- Nvidia says Anthropic, OpenAI among users of new Vera chip — Nvidia's new inference chip is already in the hands of the big labs.
Pass it on