Prates FYI

Google Today · Breaking

Google I/O 2026: Gemini Everywhere, Android XR Glasses, and the End of Chromebooks

Google's annual developer conference kicked off today in Mountain View with a Gemini-saturated keynote. The headline story isn't a single dramatic model release but rather deployment density: Gemini 3.2 Flash is rolling into Search, Maps, YouTube, Docs, Gmail, and Chrome for billions of users simultaneously. Google also previewed its Android XR smart glasses—running Gemini 2.5 Pro for real-time translation, navigation, and visual understanding—and debuted Googlebooks, premium Android laptops replacing Chromebooks entirely. On the agentic side, Gemini Intelligence lands as a system-level AI agent for multi-step task automation on Samsung and Pixel devices this summer, while Gemini Spark brings autonomous inbox decluttering, meeting briefs, and news digests to Workspace users.

        Developer note: Gemma 4, Google's open-weight model family, is expected to be announced
        alongside expanded Cloud AI APIs during Day 2 sessions on May 20.
      

→ Android Central Live Blog → Let's Data Science

Anthropic This week

Anthropic Near $1 Trillion Valuation — Would Leapfrog OpenAI

Anthropic is in advanced talks to raise between $30 billion and $50 billion in a new funding round that would value the company at up to $950 billion—narrowly eclipsing OpenAI's $852 billion valuation from March. The staggering jump in value tracks a breakout revenue moment: Claude Code has driven Anthropic to $30 billion in annual recurring revenue as of April, up from $14 billion just two months earlier. The coding agent, not the base model, has become the company's growth engine.

$950B Reported valuation target

$30B ARR (April 2026)

2× ARR growth in 60 days

→ Fortune Tech

xAI May 14

xAI Launches Grok Build — 8-Agent Coding Tool Takes Direct Aim at Claude Code

Elon Musk's xAI entered the agentic coding race with Grok Build, a command-line tool in early beta that spawns up to eight parallel sub-agents working on separate codebase branches simultaneously. The three-stage workflow (Plan → Search → Build) includes a mandatory user approval gate before any file is touched, plus Arena Mode—an automated evaluation layer that scores competing agent outputs before a developer ever reviews them. The underlying model scores 70.8% on SWE-Bench Verified with a 256K-token context window, priced at $0.20/$1.50 per million tokens for API users. Access requires a SuperGrok Heavy subscription ($299/month), though xAI is dangling a $99/month introductory SuperHeavy tier for the first six months.

→ DevOps.com → Techloy

OpenAI This week

OpenAI Ships Workspace Agents, Codex Goes Mobile and Gets a Windows Sandbox

OpenAI rolled out three interlinked updates this week. Workspace Agents lets teams create shared autonomous agents in ChatGPT that handle complex, long-running workflows—preparing reports, writing code, responding to messages—all governed by organization-set permissions. Separately, Codex arrived on the ChatGPT mobile app in preview, letting developers review work, approve commands, and steer threads from their phones. A new Codex sandbox for Windows adds firewall-backed network isolation and tighter file-write controls for enterprise teams. All three products run on GPT-5.5, which OpenAI describes as its most intuitive model to date, with HIPAA support now available for enterprise Codex users.

→ OpenAI Blog → GPT-5.5 Announcement

Cursor May 13

Cursor's Biggest Infrastructure Overhaul Since Cloud Agents: Multi-Repo, Teams, Rollback

Cursor shipped a major infrastructure update that the team says is the most significant since cloud agents launched in February. Key additions: multi-repo support, Dockerfile configuration with 70% faster layer caching, version history with rollback, scoped secrets management, and a Microsoft Teams integration. The pace of adoption is striking—more than 35% of pull requests merged by Cursor's own engineering team are now authored entirely by autonomous cloud agents, a figure the company is using as proof-of-concept for the product's enterprise roadmap.

→ PrimeAICenter

Anthropic May 14

Anthropic Closes $200M Partnership With Gates Foundation, Launches Claude for Small Business

Two product moves from Anthropic this week signal expansion beyond its enterprise core. A $200 million partnership with the Gates Foundation will deploy Claude across global health research, agricultural development, and financial inclusion programs in the developing world. Separately, Anthropic launched Claude for Small Business (May 13), a lower-priced tier with simplified onboarding aimed at companies that lack dedicated AI procurement teams—a direct response to OpenAI's aggressive SMB push.

→ Anthropic News

GitHub / Microsoft Effective June 1

GitHub Copilot Ends Flat-Rate Subscriptions — Usage-Based AI Credits Kick In June 1

Starting June 1, GitHub Copilot moves from its familiar Pro ($10/month) and Pro+ ($39/month) tiers to usage-based AI Credits billing. The change affects all personal plans; Business and Enterprise customers will see a parallel credits system. GitHub paused new Copilot Pro and Pro+ sign-ups in April ahead of the migration. The shift mirrors a broader industry trend: as model costs fall and usage patterns diverge, flat monthly subscriptions increasingly leave money on the table for heavy and light users alike.

→ Releasebot

Policy May 5

U.S. Government Gets Pre-Deployment Testing Access to Google, Microsoft, and xAI Models

The Center for AI Standards and Innovation (CAISI) announced voluntary agreements with Google DeepMind, Microsoft, and xAI that allow the U.S. government to evaluate AI models before public release. CAISI will conduct targeted safety research and pre-deployment evaluations under the deals—the first time all three companies have granted federal evaluators this level of access simultaneously. The agreements do not give the government veto power over releases, but represent a significant normalization of state-level AI oversight in the United States ahead of potential federal legislation.

→ CNBC

Tuesday, May 19, 2026