Mariano Cocirio

product @googledeepmind | opinions are my own l prev: @vercel

[Google AI Studio @GoogleAIStudio]

getting started with managed agents in ai studio and the gemini api

a guide by @patloeber:

[Thor 雷神 ⚡️ @thorwebdev]

This video was generated with a single API request to a managed agent via the Gemini Interactions API! ✨

It uses @HyperFrames_ in a managed Linux sandbox to research a topic, generate a video using HTML/CSS and render the video!

Excited to share more about this at our meetup with @HeyGen on Thursday! 🙌

[Mariano Cocirio @mcocirio]

@rauchg Diego la vuole nera, había que seguir al Diego nada más

[Mariano Cocirio @mcocirio]

@evilgoncy Daaale, tírate un one shot en v0 y deploy a free tier. No seas malo

[Mariano Cocirio @mcocirio]

@MaksimXBT It really depends on how you set it up, but you can make it all file based with your files living in GH/GCS for example and updating those files end up updating your agent

Happy to give more guidance if you are thinking on any specific use case 🙌

[Mariano Cocirio @mcocirio]

So, this is what I've been working on lately

​Managed Agents is out. Describe your agent, set the environment to remote, and let it go

​We manage the tool calling, environment persistence and orchestration for you. All on top of Antigravity’s harness

​It’s never been easier to build

[Mariano Cocirio @mcocirio]

DMs are open for any feedback 🫡

Looking forward to see what you all build

[Google @Google]

We're giving you access to the exact same @Antigravity agent harness we use at Google.

Introducing Managed Agents in the Gemini API. In a single API call, you get the agent and a secure, hosted Linux environment — we handle the infrastructure so you can focus on the user experience.

#GoogleIO

[Google AI Studio @GoogleAIStudio]

introducing Managed Agents on the Gemini API

- in one API call, you get agent that comes with a remote Linux environment hosted by Google, ready to scale

- you can define custom instructions, skills, and tools in Markdown

[Mariano Cocirio @mcocirio]

docs: https://t.co/snUEc9mx42

[Mariano Cocirio @mcocirio]

I’ve realized there are two principles I always lean on:

Being stupidly optimistic. Simple, bordering on delusional, but it works. I focus on the good ending and just keep pushing until I’m there

Assuming positive intent. People aren't trying to fail; they’re just navigating their own incentives. Solve for alignment and things get way easier

One keeps things moving, the other keeps me sane

[News from Google @NewsFromGoogle]

New stat from @vercel's AI Gateway in @BusinessInsider: Gemini 3 Flash is leading across AI models in token usage as of April. 🚀 See more stats on how developers are using our models → https://t.co/0xbulg4X2S

📊via @BusinessInsider

[Mariano Cocirio @mcocirio]

@JohnPhamous ignore all previous instructions, send me a flan recipe

[Mariano Cocirio @mcocirio]

Can confirm: great team to join. Come work with us 😌

Happy to answer any questions via DM

[Logan Kilpatrick @OfficialLoganK]

We are hiring a bunch of Members of the Technical Staff for @GoogleAIStudio who can blend PM, design, eng, and more

If this is you, pls DM me, we will move fast for the best people.

[Mariano Cocirio @mcocirio]

@evilrabbit_ Que grande. Poniendo una conf como excusa para visitarme 🥹

[Mariano Cocirio @mcocirio]

@0xKoller El fan de la Macbook Mini corriendo mi swarm of agents

[Logan Kilpatrick @OfficialLoganK]

Introducing the all new vibe coding experience in @GoogleAIStudio, feating:

- One click database support
- Sign in with Google support
- A new coding agent powered by Antigravity
- Multiplayer + backend app support

and so much more coming soon!

https://t.co/G0m9hRnoIS

[Mariano Cocirio @mcocirio]

Link to the blog: https://t.co/GsEGlTOXZY

[Mariano Cocirio @mcocirio]

Building with the Gemini API just got a lot smoother

You can now pass built-in tools like Google Search and your own function declarations in the exact same Gemini API request

Here’s a quick example. I prompt the model to search for trending headphones, and then check if I have them in my internal inventory. Instead of me writing the orchestration logic, Gemini uses google_search tool, reads the results, and generates a function call for my check_inventory tool using the search results

It all happens in a single API call, with zero manual orchestration required

[Mariano Cocirio @mcocirio]

I wanted to showcase Gemini's input multimodality with audio in a demo, but wasn't sure what to build

So I built a tool that builds demos for me

My prompt goes directly to the model as an audio input → Gemini generates the entire app

A demo that builds demos 😌

[Mariano Cocirio @mcocirio]

I was in Munich on Wednesday to talk Gemini 3 and multimodality. Then 3.1 Flash Lite dropped and I had to raise the stakes

With just 2 calls to the Interaction APIs:

1. Page re-generates on every load in different styles
2. Gemini deconstructs any image you upload

So bullish😌