Google shipped 11 AI agents this week, and the reaction wasn't what you'd expect.

Twitter memes and reactions are understandable and i wouldnt take them seriously but throughout the keynote, the audience IRL never gave an applause when the presenters paused expecting them.

And one Reddit user wrote afterward that the whole presentation had left them "repulsed at technology even though a lot of what they presented was really impressive."

It was the strangest I/O I've watched in years. Let me walk you through what they launched.

Spark is the agent that is probably the one that frames everything else Google launched this week.

It's a personal AI agent that runs on dedicated Cloud VMs at Google, which means it keeps working in the background when your laptop is closed and your phone is in your pocket.

The demo that landed for me was Spark parsing a credit card statement overnight and flagging the four subscriptions you forgot you'd signed up for.

Mundane in the abstract, but it's the kind of small task I'd happily hand off.

Right now it's in trusted testing, with an AI Ultra beta opening in the US next week.

Daily Brief is essentially Spark Lite, with the key difference that it ships today.

It runs overnight against your inbox, calendar, and tasks, then hands you a personalized digest in the morning.

My guess is that most people reading this will use Daily Brief for the next six months even if they have Ultra access, because it gives you about 70% of the Spark experience without any of the "an AI is emailing my boss" anxiety.

It's available right now to AI Plus, Pro, and Ultra subscribers in the US who have their Google apps connected.

Information Agents in Search lets you spin up multiple background agents inside Search that monitor whatever you tell them to.

That could be a topic you want updates on, a stock you're tracking, or even just a flight price you're hoping will drop.

Each agent sends back synthesized updates rather than the raw alert stream you'd otherwise be drowning in.

It's useful in theory, but it only really pays off if you check these things often enough for background monitoring to matter.

Most people don't, which means I expect this to be one of those features that sounds great in a demo and gets quietly ignored after the first week.

It's coming this summer to AI Pro and Ultra subscribers.

Universal Cart is the launch that could matter most to Google's business long-term.

It lives across Search, Gemini, YouTube, and Gmail, and the moment you add something to it, the cart starts working in the background. It watches for price drops and inventory restocks, and goes as far as flagging products that aren't compatible with the rest of what you already own.

The keynote demo caught a PC build mismatch before checkout, which is the kind of catch that saves real money.

The bigger story here is that Google now owns both the discovery layer and the purchase layer for online shopping at the same time, and that's a position Amazon should probably be watching very carefully.

Flow Agent is overdue.

Until this week, Flow could only run a single prompt at a time, which made any serious creative project painful to work on.

The new agent can plan multi-step work and generate batches of variations to compare side by side.

It can also act as a sounding board on dialogue between characters or mid-project plot adjustments.

It's already live globally for all Flow users, and if you've used Flow seriously for anything you already know why this matters.

Workspace picked up three voice-first agents at once.

Voice in office settings still feels wrong to me as a category. Nobody wants to be the person dictating an email at the open-plan desk next to two other people trying to focus.

The dictation use case has legs, but I think the rest of these will struggle to find adoption.

All three roll out this summer to AI Pro and Ultra subscribers.

Android Halo is the most interesting design choice in the entire keynote.

It's a glow at the top of your phone's status bar that shows live progress on whatever your agents are currently doing.

Every other agentic product I've seen pretends you'd never need to check on what your agent is up to, but Halo openly admits you will and gives you a passive way to do that without opening anything.

It ships with Android 17 later this year.

1. Manus Scheduled Tasks: Set the autonomous agent to run recurring jobs on a timer, like a daily news brief or weekly competitor research, with results sent to email, a file, or a connected app. A recent upgrade also lets you pick which agent runs each scheduled task, so you can control cost and reliability on high-volume workflows.

2. Polarity: Sandboxed eval infrastructure that runs each AI agent task inside an isolated Docker sandbox loaded with real backing services like Postgres, Redis and S3, then scores the run against behavioral rules.

3. LobeHub: A Chief Agent Operator: one workspace where you hire, schedule, and run a whole team of AI agents that operate around the clock and report back without you babysitting them. It grew out of the open-source LobeChat framework, so it spans providers like GPT, Claude, Gemini and DeepSeek, with shared context, scheduled tasks, and project organization built in.

Antigravity 2.0

Google's new agent-first IDE, and it's now a standalone desktop application rather than an extension you bolt onto something else.

You can run multiple agents in parallel inside it, each in its own sandboxed terminal with credential masking.

The agent harness underneath all of this is the same one Google's internal teams have been using to ship product features all year.

The translation is that Google is taking this category seriously in a way they haven't taken IDEs before.

It's worth installing if you're already in the Gemini ecosystem.

Managed Agents in the Gemini API

This is the most useful thing Google shipped this week if you're a developer.

One API call gives you a remote Linux sandbox where the agent can reason about a task, run code, manage files in isolation, and browse the web for live data.

You define what the agent knows how to do in markdown files like AGENTS.md and SKILL.md, which removes most of the orchestration code you'd otherwise have had to write yourself.

If you've been blocked on shipping an agent because you didn't want to stand up the infrastructure, that blocker just went away.

Android Migration Agent

The Android Migration Agent ports your React Native, web, or iOS codebase into a native Kotlin Android app.

Google's claim is that work that used to take weeks now takes a few hours, which sounds like marketing copy at first read, but if it's even half-true it's worth testing on something low-stakes.

It's currently in preview inside Android Studio.

The chatbot was a transitional UI, and the next interface doesn't ask to be opened at all.

Your agent is already running in the background, doing things you'd otherwise be doing yourself, and surfacing what matters when it matters.

That's the bet Google is making.

Whether you find it useful or invasive comes down to one question: how much of your inbox, calendar, photos, and shopping behavior are you willing to hand over to make all of this work?

Because none of these agents work without that level of access.

Flash beating the previous Pro on Google's own benchmarks is nice, and Spark could shift how people use their phones if it works in practice the way the demos suggested it does.

Universal Cart could rewire a meaningful chunk of online shopping, and the eyewear partnership with Samsung and Warby Parker is the most credible smart glasses play since Ray-Ban Meta started selling out.

But the show itself was strange.

Pichai walked off at the 43-minute mark and never came back.

Demis Hassabis closed the whole thing with a line about "standing in the foothills of the singularity," delivered into flat silence.

Nobody in the room seemed to know how to react to what they were being shown.

I think the problem is with how the whole thing was framed.

Google spent the keynote showing token charts as if 3.2 quadrillion tokens processed last month was supposed to mean something to anyone outside the AI industry, which it doesn't.

People wanted to see Spark do something useful on a Tuesday afternoon, and what they got was an architecture diagram.

Apple's WWDC is two weeks away, and Apple usually wins on framing even when they lose on specs.

If Cupertino is paying attention, they'll spend their keynote on a real person using a real feature instead of bragging about how the feature got built.

I'll wait for Spark to hit public beta and run it for a week before I commit to anything.

But I like where it’s going.

The next AI interface probably won't be one you have to open, and Google just made a poor case for it this week.

Until next time,
Vaibhav 🤝🏻

If you read till here, you might find this interesting

#AD 1

You don't send a colleague a three-word brief. You explain the context, the constraints, what you've already tried. But typing all that into ChatGPT takes forever — so you don't.

Wispr Flow lets you speak your prompts instead. Talk through your thinking naturally and get clean, paste-ready text. No filler words. No cleanup. Just detailed prompts that actually get you useful answers on the first try.

Millions of users worldwide. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free

#AD 2

Your CFO opens Slack. There's a weekly Stripe revenue recap in #finance with a churned-accounts flag and a net-new breakdown. She didn't ask for it.

Your head of product opens Slack. There's a GitHub summary in private channel: PRs merged, PRs stale, Linear tickets that moved. He didn't ask for it.

Your marketing lead opens Slack. There's a Google Ads performance comparison in private channel, with a note: "Meta CPA crept up 18% this week. Might be worth pausing the broad match campaign." She didn't ask for it either.

All-hands at 10am. Everyone already knows the numbers. The meeting is about decisions, not catch-up.

That's what happens when one colleague works across every tool your company uses. Not one department's assistant. The whole company's coworker.

Viktor lives in Slack. Top 5 on Product Hunt, 130 comments. SOC 2 certified. Your data never trains models.

"Not only have we caught up on several months of work, we are automating manual tasks and expanding our operations to things previously not possible at scale." - Jesse Guarino, Director, Torque King 4x4

Start free. $100 in credits →