Sunday May 24
PRODUCTGoogle

Google Picks Its Fight: It's The Stack

19MAY

Google I/O ships Gemini 3.5 Flash (now generally available), Spark (a 24/7 personal AI assistant), Antigravity 2.0 (an agent-building platform), and Android XR smart glasses.

Google stopped competing on best model. Started competing on best agent platform. The advantage: surfaces OpenAI and Anthropic can't reach. Android. Chrome. Workspace. Cloud.

The procurement question shifts from "which model?" to "where do my agents run?" Antigravity 2.0 is now a five-surface stack, not just an IDE. Most enterprises will end up multi-platform within 18 months whether they planned for it or not.

Spark gated behind Ultra means cost-per-request hasn't hit consumer economics yet. No Gemini 4.0 is conspicuous. Google chose platform over leaderboard this quarter.

full brief & sources

⚡ Why this matters

  • Google chose to compete on agent platform, not model leaderboard.
  • Advantage: surfaces (Android, Chrome, Workspace, Cloud) that OpenAI and Anthropic can't reach.
  • Procurement question shifts from "which model is best?" to "where do my agents need to run?"

🔍 What happened

  • May 19, 2026, 10am PT. Google I/O at Shoreline Amphitheatre.
  • Gemini 3.5 Flash GA across products and API. More expensive than 3.0. Plan-everything default.
  • Gemini 3.5 Pro arrives next month. No Gemini 4.0 this quarter.
  • Gemini Spark: 24/7 agentic personal assistant built on 3.5 + Antigravity. Cloud-resident. Ultra-only next week.
  • Antigravity 2.0: five-surface stack (Desktop App, agy CLI, SDK, Managed Agents API, Enterprise Agent Platform).
  • agy CLI replaces the deprecated Gemini CLI.
  • Android XR Glasses confirmed with Samsung, Warby Parker, Gentle Monster, XREAL. Fall 2026 release.
  • Rocky rollout: Antigravity 2.0 auto-update wiped local configs and removed built-in code editor.

💬 Smart takes

  • Simon Willison: Flash 3.5 is meaningfully more expensive than 3.0. Google is consolidating around fewer, more capable models per tier.
  • Ben Thompson: DeepMind alignment with Google's business objectives is the open question. Spark execution lagged research and platform announcements.
  • Skeptic: Rocky Antigravity 2.0 rollout costs developer trust. Spark gated behind Ultra means cost-per-request hasn't hit consumer economics. No Gemini 4.0 is conspicuous given Anthropic's $950B framing.

🧭 Where this goes

  1. "Agent stack" becomes the procurement category by Q4. Enterprises evaluate Antigravity vs Claude Code Enterprise + Cowork vs Codex + Operator + DeployCo.
  2. Spark is the consumer-facing test. If Ultra subscribers actually use it without cost economics blowing up, agentic-as-product crosses the consumer threshold.
  3. Antigravity 2.0 stabilization decides developer mindshare in the next 30 days.
  4. Android XR Glasses at sub-$500 in fall 2026 makes spatial AI procurement-relevant.
  5. Gemini 4.0 silence becomes a Q3 announcement window.

🎯 Implication

  • For PMs evaluating AI vendors: stop scoring on benchmarks. Score on surfaces your agents need to run in. If your product lives in Android, Chrome, Workspace, or Google Cloud, Antigravity 2.0 is a serious option.
  • For execs: most enterprises will end up multi-platform within 18 months whether they planned for it or not.
·TechCrunch - With Gemini 3.5 Flash, Google bets next AI wave on agents not chatbots (May 19)·TechCrunch - Gemini Spark, 24/7 agentic assistant (May 19)·TechCrunch - Antigravity 2.0 desktop + CLI (May 19)·Simon Willison - Gemini 3.5 Flash analysis (May 19)