NEW Antigravity Agent Teams Update is INSANE! artwork

NEW Antigravity Agent Teams Update is INSANE!

AI News Today | Julian Goldie Podcast

July 16, 2026

NEW Antigravity Agent Teams Update is INSANE!
Speakers: Julian Goldie
**Julian Goldie** (0:00)
There's a brand new update from Anti-Gravity. It's called Agent Teams in Anti-Gravity. This is a new powerful system to build a dynamic team of specialized sub-agents they can coordinate in the background to plan, build, and verify complex systems in the background of working parallel, which is really, really powerful. So to do this, we've actually plugged in the Anti-Gravity CLI into our Agent OS, as you can see right here. And just to test it out, I actually ran teamwork, I ran a preview for an SEO blog as a test over here. And you can see that actually came up with the answer very quickly. So it creates a markdown file with a full plan for your agents. And then you can actually run it live, as you can see right here, and spin up the agentic sprint directly in the dashboard, which is pretty powerful. So you can just open up and start running it straight away. And then the cool thing about this is the whole plan gets saved inside your workspace, so we can come back to it later. So this is something I call the One Command Company. So Google's anti-gravity can now spin up a whole team of AI specialists from One Command. They plan, they build, they check in parallel, and hand you the finished thing. I actually tested it today, and two of the builds are playable right inside the guide that I'm going to show you. Now you might wonder, okay, why is this useful? So I will talk about the Q of One problem, which is, you know, imagine running a company with a single employee. Now, brilliant, but it's still just one. They plan the project, they build the first part, then the second part, then they check it one thing at a time and in line. Now, if you double the amount of tasks they have to do, you double the weight. And who splits the work into pieces? Who writes each job description? Who remembers to check the result before it ships? Well, you do, on every project, every single time. And that's what working with one AI agent is like. The model is brilliant, but the project management never stopped being your job. The one command company set up with anti-gravity breaks that cycle because one command basically hires a team, briefs the team and runs the team as well.
So what does agent teams actually do? So anti-gravity is Google's AI coding agent. It's the successor to that GeminiC like. And until now, it was one worker. However, when you use a type and you type forward slash teamwork hyphen preview, it becomes a hiring manager. So you get a coordinator that reads your task, then drafts a team sheet, which specialists to hire, what each one owns and how they'll hand the work to each other. You say, yes, it spins the team up for real. So you literally just approve the plan, and then it will build the whole thing out once, which is pretty incredible.
So the way this works is you have a coordinator that drafts a team and reads the work, the plan designs, builder A builds, builder B builds, and you get a verifier checker that works through it as well. So here's an example. We actually created this plan for the team sheet as I showed you a second ago. So inside our HNOS, we have anti-gravity over here, and then inside the chat, we said, teamwork preview, create that page. And so what it actually does is it creates an idea of team sheets here. So we've got the SEO architects, the next JS developer, SEO Ops, and the LighthouseX, which is a QA developer as well. It actually plans out exactly how it's going to be built. And then the same run can actually be simulated live. And here's an example of what it can look like in real time. So we have the multi-agent collaborator preview, the preview orchestrator. We can switch between sub-agents and the template presets. Then it has the full project objective features. And then it goes off and starts spinning up the full workspace with your team of agents working together, which is pretty awesome. Now, some people say you need to be technical to run an AI team of agents. Actually, no, because literally all you do is in one single command, you type forward slash teamwork hyphen preview, and you're good to go. Now, other people say, well, more agents must get more expensive. Actually, you can use the CLI for free.
Or you can actually plug in your own custom APIs into there as well. You could even, for example, run this with a local model like Gemafor, which comes from Google directly as well. And then other people say, well, it's a preview. I should wait. I don't want to start using this system straight away. But honestly, the people who start building right now are going to get ahead. The people who wait are going to fall behind. And so you can see some examples of what our team actually built here in terms of how they've worked together. So with the builder, the planner, the coordinator, and with the verifier too. And then we can see the tasks that are being worked on.

6 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000777130718