Build · founder · 7 min read

Google I/O 2026 for Founders: What Actually Matters

Antigravity 2.0, Android vibe coding, Gemini 3.5 Flash, and the Gemini CLI sunset — here's what changed and what you should do about it.

Published May 19, 2026 ·

Google IO Antigravity AI coding Gemini Google AI Studio

Google I/O ran today. If you watched the full keynote, you saw a lot of demos involving robots, Android XR glasses, and Gemini doing impressive things in Search. Almost none of that is relevant to you right now.

Here’s what actually matters for non-technical founders and product builders who are in the middle of shipping something with AI tools.

The biggest news: Antigravity 2.0

If you’ve been following the vibe coding space, you know Google launched Antigravity as its “agent-first” IDE in March 2026. The first version was technically impressive but rough — developer-native, Firebase-dependent, and not really designed for anyone without a codebase already open on their second monitor.

Version 2.0, announced today, is a different product.

The desktop app got rebuilt. It’s now a proper standalone application rather than a tab inside Google AI Studio. The interface is built around what Google calls “Mission Control” — a dashboard where you can dispatch multiple agents to work on different tasks simultaneously and watch their progress from a single view. One agent can be writing a new feature while another runs tests. A third can be reviewing a pull request. You see all of it in real time.

There’s a built-in browser. Google builds Chrome, and it now lives inside Antigravity. Your agents can interact with your running app — click around, take screenshots, fill forms — without any setup. Testing a user flow no longer requires you to manually babysit the loop between “code change” and “did that actually work in the UI.”

Voice commands. You can describe what you want without typing. This is more useful than it sounds for the kind of rapid iteration where you’re staring at a screen and thinking out loud.

New pricing. Antigravity exited preview pricing. The free tier continues with rate limits. Pro is $20/month. A new AI Ultra tier at $100/month gets you 5x the usage limits and priority model access — relevant for teams running heavy parallel workloads. For the I/O week (until May 25), Ultra comes with a $100 bonus credit offer.

Who Antigravity 2.0 is for

Still not non-technical founders. The interface is still fundamentally an IDE with a task manager layered on top. If you can’t read a diff and don’t know what a git branch is, the Mission Control dashboard will confuse rather than help.

But if you’re a technical founder, a PM with a technical co-founder, or someone who’s already comfortable with tools like Cursor — this is the most capable option Google has shipped. The parallel agent architecture, the Firebase backend, and the Chrome integration are a genuine combination that doesn’t exist elsewhere.

Gemini CLI is being discontinued

This one matters if you or anyone on your team has been using Gemini CLI.

As part of the Antigravity 2.0 launch, Google announced that Gemini CLI is being absorbed into Antigravity CLI — a new CLI tool built in Go, faster, and designed for multi-agent workflows. The old Gemini CLI will stop serving requests for Google AI Pro and Ultra subscribers on June 18, 2026. That’s about 30 days from now.

If you’re on the free tier or using Gemini Code Assist through an enterprise license, your access continues unchanged for now. But the writing is on the wall: Antigravity CLI is the future and Gemini CLI is the past.

What to do: Switch now, before the deadline. Antigravity CLI preserves the core features (agent skills, hooks, subagents) and adds multi-agent capability. The migration is designed to be low-friction — your existing workflows carry over. No reason to wait until June 17.

Google AI Studio can now build Android apps

This is the update most relevant to founders who’ve been using Google AI Studio as a free alternative to Lovable.

Until today, AI Studio built web apps — React or Next.js frontends, Firebase backends, deployed to the browser. From today, you can also vibe code native Android apps in Kotlin, using Jetpack Compose, directly inside AI Studio.

The workflow: describe your app, AI Studio generates the Kotlin code, you preview it on an Android Emulator in-browser, install it on a physical device, and publish to Google Play’s Internal Test Track — all without leaving AI Studio. When you want fuller controls, you hand the project off to Android Studio.

The honest take: This feature is for founders with a basic technical literacy around mobile apps. If you know the difference between a native app and a web app, you’ll understand when to use this. If you’re building your first app and just need something to validate an idea, the web app workflow is still simpler — a browser URL you can share is easier to test than an APK.

But for anyone who’s been holding off on building an Android version of their product because the mobile-native stack felt too technical: this lowers that barrier meaningfully.

Gemini 3.5 Flash: faster, cheaper, still frontier-quality

Google’s new flagship model launched today. The headline claims are impressive: outperforms Gemini 3.1 Pro on coding and reasoning benchmarks, at Flash pricing and 4x the output speed of comparable frontier models. It’s rolling out in Antigravity 2.0, Google AI Studio, and via the Gemini API.

What this means practically: the underlying model powering all of these tools just got better. If you’ve found AI Studio or Antigravity’s output quality inconsistent, the 3.5 Flash upgrade should improve the reliability of code generation and multi-step reasoning.

You don’t need to do anything to get it — the upgrade is automatic for existing users.

What I/O didn’t change

A few things worth naming that didn’t move today:

Lovable, Bolt, Cursor, and Replit are still the right first tools for non-technical founders. Google’s launches are impressive but they remain developer-native. The on-ramp for a first-time builder is still smoother on Lovable.

Firebase lock-in is still a real trade. If you build on AI Studio or Antigravity, your backend is Firebase. That’s a fine stack — mature, scalable, well-documented — but it’s a different trade-off than the Supabase-and-Vercel stack Lovable produces. Neither is wrong; they’re just different bets.

Android I/O demos are more impressive than daily-use products. Google’s live demos always pop. The real question is what the daily editing experience feels like in week three of a project, not in a keynote where the happy path was pre-scripted.

The one-line summary

Antigravity 2.0 is now a serious IDE for technical founders. Gemini CLI users have 30 days to migrate. Google AI Studio added Android. Gemini 3.5 Flash made everything faster. Everything else is noise.