DAILY INTELLIGENCE BRIEF

The Vault - AI Edition

Monday, June 22, 2026
16 stories
15,908 words
68 min read

Business

4
01

Noam Shazeer, Transformer Co-Author, Leaves Google for OpenAI as Lead for Architecture Research

Noam Shazeer, the engineer whose ideas sit at the foundation of nearly every modern large language model, is leaving Google.

4 minRead →
02

SpaceX Plans to Build AI Data Centers in Orbit as Terrestrial Power Runs Short

Ten days after pulling off the largest IPO in history, SpaceX is asking investors to believe in space-based compute.

4 minRead →
03

Epic's Unreal Engine 6 Makes Claude and Gemini Core Engine Pillars

When Epic Games walked onstage at Unreal Fest Chicago on June 17, 2026, the most consequential announcement was AI.

5 minRead →
04

OpenAI and Anthropic File Confidential IPOs as the AI Listing Supercycle Arrives

In the span of eleven days, the two companies that defined the generative-AI era both told Wall Street they want to go public.

4 minRead →

Llms Genai

4
05

Google's Android 17 Ships With Gemini Omni and Lyria 3 Music Generation Built In

Google has stopped treating AI as an app you open and started treating it as part of the operating system itself.

4 minRead →
06

OpenRouter's Fusion Blends Multiple Models to Rival Fable 5 — at Half the Cost

The day after Anthropic's flagship Fable 5 went dark under a US export-control order, OpenRouter shipped a product.

4 minRead →
07

Claude Code Ships Nested Subagents to Tame Long, Multi-Hour Coding Sessions

For most of the past year, the dirty secret of agentic coding has been the same one that has dogged LLMs: context.

4 minRead →
08

OpenAI's Record and Replay Turns One-Off Workflows Into Reusable Codex Skills

OpenAI has quietly handed Codex a new way to learn: just show it what to do, once.

4 minRead →

Research

4
09

Anthropic Study of 400,000 Sessions Finds Domain Expertise Beats Coding Skill in Claude Code

The conventional wisdom about coding assistants has been that the better a programmer you are, the more you get out.

5 minRead →
10

AI Agents Resolve Open Math Conjectures With Proofs Checked Line-by-Line in Lean 4

A pair of AI agents has resolved a mathematical question that sat open for more than a decade — and proved it formally.

4 minRead →
11

Google DeepMind's AI Co-Mathematician Scores 48% on FrontierMath's Hardest Tier

Google DeepMind has built a research assistant for mathematicians that sits beside them rather than just answering.

4 minRead →
12

KAIST's VOTP Teaches AI Human Judgment From a Handful of Preference Videos

Teaching a robot to act the way a person would want has long carried a hidden tax: someone has to label the data.

4 minRead →

Policy

4
13

42 State Attorneys General Launch Sweeping Investigation Into OpenAI Ahead of Its IPO

A coalition of 42 state attorneys general has opened a sweeping investigation into OpenAI, with subpoenas in hand.

4 minRead →
14

China Unveils a $295 Billion Five-Year AI Infrastructure Plan

Beijing has put a number on its ambition: a $295 billion, five-year plan to fund a nationwide AI buildout.

4 minRead →
15

The White House Executive Order That Explains the Fable 5 Ban

When the U.S. suspended foreign access to Anthropic's most powerful AI models, the order did not come out of nowhere.

5 minRead →
16

Congress Unveils the 'Great American AI Act,' Its Most Comprehensive Federal AI Framework Yet

Two House lawmakers from opposite parties have produced the most ambitious attempt yet at a national AI rulebook.

5 minRead →