Daily AI Briefing — June 2, 2026

The AI news cycle is relentless. Microsoft Build 2026 opens today with its biggest AI push yet — in-house models to reduce OpenAI dependency. OpenAI's Codex crossed 5 million weekly users and is now targeting knowledge workers. Anthropic filed for IPO. And Berkshire just bet $10 billion on Alphabet's AI infrastructure buildout. Here's the briefing for June 2.

Microsoft Build 2026 Kicks Off — MAI In-House Coding Model for GitHub Copilot

Microsoft opens its Build 2026 developer conference in San Francisco today with the launch of a new family of in-house AI models collectively branded MAI, including a specialized coding model engineered to power GitHub Copilot. This is a strategic shift: Microsoft is reducing its reliance on OpenAI by developing task-specific models — coding, reasoning, and lightweight — built in-house. The coding model is expected to improve Copilot's autonomous capabilities and multi-file editing, areas where Copilot has lagged behind Cursor and Claude Code. Source: Windows News / Seeking Alpha via llm-stats.com

For GitHub Copilot users, this is the biggest update since agent mode launched. If Microsoft's in-house coding model delivers on cross-file reasoning, it closes Copilot's biggest weakness — and at potentially lower cost since it eliminates OpenAI licensing fees. The MAI family signals Microsoft's ambition to own the full AI stack, from model to IDE.

OpenAI Codex Hits 5M Weekly Active Users — Expanding into Knowledge Work

OpenAI released its report "The Next Era of Knowledge Work", revealing Codex now has 5+ million weekly active users — up more than 6× since the desktop app launched in February. Knowledge workers now represent about 20% of users and are growing more than three times as fast as developers. The fastest-growing tasks are data analysis (110% week-over-week), research (37%), and knowledge artifact creation (36%). Source: OpenAI official report

The key insight: ~50% of Codex users now keep more than one task running simultaneously — inspecting a dataset on one thread, drafting a script on another, assembling a report on a third. Codex is evolving from a coding tool into a general-purpose task automation platform. This directly benefits the Codex CLI ecosystem, which provides terminal-native access to the same underlying agent infrastructure.

Anthropic Files for IPO — Race to Go Public Before OpenAI

Anthropic has confidentially filed for an initial public offering, per SEC disclosures. The filing comes days after the company closed a $65 billion Series H at a $965 billion valuation, surpassing OpenAI as the world's most valuable AI startup. The IPO race with OpenAI — which filed its S-1 in April — is now fully engaged, with both companies racing to set the financial reporting template for frontier AI. Source: New York Times

This is a market structure event: two trillion-dollar AI companies trying to IPO in the same window, alongside SpaceX's rumored $75B listing. The capital market disruption could be significant. For users of Anthropic's Claude Code and Claude Opus 4.8, the IPO signals that Anthropic's enterprise focus is deepening — expect more compliance features and enterprise sales motion post-IPO.

Berkshire Hathaway Invests $10B in Alphabet's AI Infrastructure

Berkshire Hathaway will purchase $10 billion of Alphabet stock through a private transaction as part of Alphabet's broader $80 billion equity raise to fund AI compute infrastructure. Alphabet's 2026 capital spending is expected to hit $190 billion. Berkshire's investment is a major validation of Alphabet's AI strategy from Warren Buffett's traditionally tech-cautious firm. Source: CNBC

The scale is staggering: Google Antigravity, Gemini, and Google Cloud's entire AI product line will benefit from this capital injection. For developers building on Google's AI platform, this means more capacity, lower pricing pressure, and faster product iteration ahead.

Alibaba Launches Qwen3.7-Plus — Multimodal Agent with Vision and Tool Invocation

Alibaba's Qwen team released Qwen3.7-Plus on the Bailian platform, a multimodal model adding image and video understanding, deep reasoning, and tool invocation. The model features self-programming capabilities and autonomous iteration — it can write its own tools to solve tasks. Alibaba's shares rose 6.84% on the announcement. Source: MarkTechPost

Qwen3.7-Plus continues the pattern of Chinese AI labs pushing multimodal agent capabilities at competitive price points. It joins a crowded field alongside DeepSeek V4 Flash and Kimi K2.6 in the open-weights agent tier.

Nvidia Open-Sources Enterprise Agent Toolkit — NemoClaw Framework

Nvidia launched an open-source enterprise AI agent toolkit including the NemoClaw framework — a security-hardened agent orchestration stack — and CUDA-X Agent Skills offering plug-and-play capabilities for data analytics, scientific simulations, and quantum computing. The company also unveiled RTX Spark, its first laptop chip purpose-built for running AI agents locally. Source: Dataconomy

Nvidia is stacking the entire enterprise AI agent stack: silicon (RTX Spark), orchestration (NemoClaw), skills (CUDA-X), and security. For teams evaluating NemoClaw, this means deeper GPU optimization and first-class CUDA support. The RTX Spark chip is particularly significant — it's the first time running sophisticated multi-agent systems on a laptop is practical without cloud offload.

xAI Enters the Coding Agent Race with Grok Build

xAI unveiled Grok Build, its first coding agent, now in early testing on a waitlist. Key differentiators: runs up to 8 parallel AI agents through a plan → search → build workflow, a unique Arena Mode that automatically evaluates and ranks competing outputs, and local-first execution — no source code transmitted to xAI servers. Powered by the custom grok-code-fast-1 model (70.8% SWE-bench Verified) at $0.20/M input tokens. Source: DevOps.com

Grok Build brings genuinely novel ideas to the coding agent space — particularly Arena Mode for automated code review and local-first privacy for regulated industries. But with only 256K context window vs. Claude Code's 1M tokens, and a waitlist that Musk teased weeks ago without shipping, it's still early days. The coding agent race now has three major contenders: Claude Code, Codex CLI, and Grok Build — plus a dozen open-source alternatives.

Quick Hits

OpenAI ChatGPT career platform: OpenAI is turning ChatGPT into a career platform with job search integration (Indeed, Upwork, Appcast) and a CV editor — initially US-only.

SK Hynix to double memory chip capacity: The memory maker announced a 5-year plan to double production, warning shortages could last through 2030 as AI demand for HBM explodes.

Trump nixed AI EO on May 21: Per Wired, Trump revoked the Biden-era AI executive order. US officials are in chaotic talks over how to replace it, with early model access for testing being the most contentious issue.

Alphabet to raise $80B: The Google parent is raising equity capital at an unprecedented scale, with Berkshire Hathaway's $10B anchor being the headline.

Tomorrow's Watch

Track Microsoft Build 2026 keynotes for MAI coding model benchmarks and GitHub Copilot integration details. Watch for independent Qwen3.7-Plus evaluations. The Anthropic IPO filing details should emerge as the SEC processes the confidential filing. And check whether Grok Build actually ships this week — Musk's "next week" promise is now overdue.

📊 See how these tools compare →

← Back to all posts