Presented by
Good morning, AI enthusiasts. Google just turned its annual developer conference into the most product-dense AI event of the year, dropping a 24/7 background agent with its own virtual machine, a model faster than its own Pro tier, a video generation tool, and smart glasses, all in one keynote.
In today's recap:
Google's biggest I/O: Spark, Flash, and Antigravity 2.0
Karpathy joins Anthropic's frontier research
Build an agentic workflow with Antigravity CLI
Google's AI smart glasses launch this fall
4 new AI tools, prompts, and more
Google ships Gemini Flash, Spark, and Antigravity 2.0
Recaply: Google just shipped Gemini 3.5 Flash and Gemini Spark at I/O, with Flash matching 3.1 Pro on benchmarks while being 12x faster, and Spark running as a 24/7 background AI agent on its own dedicated virtual machine.
Key details:
Gemini 3.5 Flash is live on Antigravity 2.0 with a new CLI and real-time voice, letting developers build and test agentic workflows 12x faster than Gemini 3.1 Pro.
Spark runs on a dedicated VM with MCP and Connector support; Antigravity 2.0 is 12x faster than Gemini 3.1 Pro, now available globally for all users.
Gemini Omni Flash also launched on Gemini and Google Flow, with Omni Pro coming soon; Antigravity-powered generative UI rolls out to Google Search for all users this summer.
Antigravity 2.0 is available to everyone globally now; Spark begins rolling out to trusted testers this week and reaches Ultra plan users in the US next week.
Why it matters: Google I/O is where the company shows how far ahead of schedule it's willing to run. Shipping a background agent with its own VM, a faster Flash model, video generation, and a global IDE upgrade in one keynote isn't just a product drop. It's a signal that Google wants to own the default platform for AI development. With Antigravity going global and search getting generative UI, most developers will end up inside Google's stack whether they planned to or not.
PRESENTED BY MODA
Moda is the AI design agent with taste
Moda's viral launch hit 4.4 million views in two days. Tens of thousands of professionals signed up. Startups, agencies, forward-thinking brands and top firms are now using Moda to create brand-aligned slides, ad creative, reports, social carousels and more.
Most AI tools tend to create what we call "AI slop": repetitions of the same colors, layouts and fonts. And when you try to fix it, you get stuck in a loop of re-prompting.
Moda is different. Drop in your website URL, and Moda learns your brand from the ground up: your colors, your fonts, your visual language. Then it helps you generate pro-quality slides, docs, and marketing assets.
The best part? Every layer is fully editable on a real canvas, and exports to powerpoint, PDF and more.
ANTHROPIC
Karpathy joins Anthropic, returns to frontier research
Recaply: Karpathy just announced that he is joining Anthropic R&D team. Karpathy said the next few years at the frontier of LLMs will be "especially formative."
Key details:
Karpathy joins to do core LLM research at Anthropic, leaving behind Eureka Labs, his own AI education startup.
He co-founded OpenAI in 2015, ran Tesla's Autopilot team, and launched Eureka Labs in 2024. He's one of the most recognized names in deep learning.
Karpathy said the next few years will be "especially formative." He chose Anthropic over OpenAI, Meta, and Google to be part of it.
He starts immediately; Anthropic reached a $380B valuation in early 2026, placing it among the three most valuable private AI companies.
Why it matters: Karpathy doesn't take jobs for status or money. His last move was building Eureka Labs himself because he thought AI education was the most important problem to solve. Picking Anthropic over OpenAI or Meta is a real statement. He's betting that safety-first research is where the most important work is happening right now. His track record makes that bet worth watching.
GUIDES
Build your first agentic workflow with Antigravity CLI
Recaply: In this tutorial, you will learn how to install Antigravity and set up your first agentic coding workflow, dispatching AI agents to plan, write, and review code with minimal input from you.
Step-by-step:
Install Antigravity from antigravity.google/download. Choose your OS, run the installer, and pick "Review-driven development" as your agent mode. It's the recommended setting: the agent works autonomously but pauses at key steps for your approval. Install the
agyCLI when prompted to open any project from your terminal.Open your project with
agy .in any code folder, or select "Open Folder" from the Agent Manager screen. This attaches your codebase to the agent's working context.Give the agent a high-level task in Agent Manager. Write something like "Refactor the auth module to use JWT tokens and add test coverage." The agent will create a planning document, list the steps it plans to take, and wait for your go-ahead before touching any files.
Review and approve each checkpoint. In review-driven mode, the agent surfaces its implementation plan and pauses before every major code change. Read the task list, approve or redirect, and let it continue. This is your safety net and it's much easier than undoing changes after the fact.
Add a workflow for recurring tasks. Go to Rules and Workflows in settings and describe a repeatable process, for example "Before every commit, run tests and check for type errors." The agent follows this automatically on all future tasks without needing a prompt.
Pro tip: Use @filename in Agent Manager to point the agent at specific files or folders. Giving it a README, a schema file, or one key module helps it scope the work correctly instead of exploring the whole project.
TOGETHER WITH TIGER DATA
Your Analytics Stack Is One Database Too Many
Pipelines, backfills, sync lag, data drift… that's the cost of splitting your stack. Tiger Cloud extends Postgres, fully managed, so analytics run on live data. No second system. Stay on Postgres. Scale on Postgres.Try Tiger Cloud free.
Google's AI smart Glasses launch this fall
Recaply: Google just revealed intelligent eyewear launching this fall through Android XR. It's partnering with Gentle Monster and Warby Parker on designs that put Gemini on your face for hands-free directions, texts, and app control.
Key details:
Say "Hey Google" or tap the frame to get Gemini for navigation, texts, calls, photo capture, and real-time speech translation, all without taking out your phone.
Two designer collections from Gentle Monster and Warby Parker launch this fall, with both iOS and Android support, built on Android XR with Samsung and Qualcomm hardware.
A second display glass type is also in development; Google XR VP Shahram Izadi said the built-in Nano Banana feature lets you say "take a picture and put everyone in funny hats" and it works.
Audio glasses launch in fall 2026 as part of the Gentle Monster and Warby Parker collections; display glasses follow in the coming months.
Why it matters: Google Glass launched in 2013 and became a joke. The camera made people uncomfortable, the design stood out in the wrong way, and the use cases felt made up. This time the approach is different: two real fashion brands, no camera anxiety since these are audio-first, a short list of daily tasks that actually matter, and Gemini as the brain. If the execution holds up, Google doesn't just have a product. It has the strongest case for AI wearables anyone has made yet.
TOOLS
Trending AI Tools
🤖 Gemini Spark - Google's 24/7 background AI agent
🧹 Remove-AI-Watermarks - Open-source Python CLI that removes AI watermarks including SynthID from images
📱 Google AI Studio (Mobile) - Google's AI Studio coming to mobile, build custom AI tools from your phone by describing what you want
🧠 MiniMax M2.7 - Runs inside Claude Code via API redirect at $40/month
NEWS
What Matters in AI Right Now?
GitHub announced it's investigating unauthorized access to its internal repositories, saying there's no current evidence of impact to customer data stored outside those systems.
Railway published an incident report showing Google Cloud incorrectly suspended its production account, causing an 8-hour platform-wide outage that took all workloads offline across every region.
Automated Tire Inc. introduced SmartBay, a $4,900/month robotic tire-change platform that processes 24 tires per hour compared to 4 for a human, with one technician able to manage up to 3 bays at once.
OpenEvidence handled nearly 27 million clinical encounters in April alone, with about 65% of US doctors actively using it and most patients unaware the free AI medical tool exists.
OpenAI announced it will embed Google DeepMind's SynthID invisible watermark into all ChatGPT, Codex, and API-generated images, layering it on top of existing C2PA metadata for more resilient content provenance.
DeepMind CEO Demis Hassabis closed Google I/O's keynote saying humanity is "standing in the foothills of the singularity," predicting AI will usher in a new golden age of scientific discovery and progress.
Pope Leo XIV and Anthropic co-founder Christopher Olah will launch the Vatican's AI encyclical "Magnifica Humanitas" on May 25, a document on human dignity in the AI era signed by the Pope on May 15.
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
