←
JY824 posts
JY avatar

JY

@agenticjay

Creative tech leading demo lab for Gemini model launches. Inspiring developers to build with Gemini at @googledeepmind @googleaistudio @antigravity

◷ Joined February 2009▧ 17 Photos and videos
JY avatar
JY@agenticjay·Nov 10

I took @simonw's benchmark “pelican riding a bicycle” way too far using Gemini 4 Carbon and a /swarm in @antigravity to build a playable game. Multi-agent critique loops ran across 5 waves to get this pelican biking across our Mountain View campus. What's most interesting is the approach they took. The subagents generated Nano Banana and Omni references to rig the assets in headless Blender before wiring the Three.js bike physics, creating custom WebGL2 shaders, and finally produced procedural Web Audio SFX with a Lyria track. Try it out on the playable link in the reply 👇

1:17
JY avatar
JY@agenticjay·11h

Get it!

make it yours: introducing custom domains in ai studio turn your idea into a live, branded app with a custom domain, all within ai studio build ai.studio/build to get started

0:30
JY avatar
JY@agenticjay·Nov 17

Yap2App

1:32
Lukas Petersson avatar
⟳ JY retweeted
Lukas Petersson@lukaspet·Nov 16

The people have spoken! Gemini 4 Carbon is the best AI at 3D printing memes. Go and vote for your favorite prints in Wojak-Bench 3D, link below.

Post attachment
JY avatar
JY@agenticjay·Nov 16

Multimodal decisions... try it on the API

Our decisions API supports multimodal input across text, images, audio, video, and PDFs. It is the only one on the market that supports this. It is super powerful, check out the real-time call router below powered by the Gemini Decisions API.

0:32
JY avatar
JY@agenticjay·Nov 16

Going to bed sending a big plan to Carbon that will run overnight then waking up to iterate on it with Hydrogen is becoming such a vibe. As part of my /swarms in @antigravity I always ask for one of the subagents to report on progress every few loops sending me short screen recordings and gifs in chat. 1) let’s me track the teams progress 2) gives me instantly shareable assets for feedback from others 3) allows me to nudge mid loops if things are going off track This is Omni Renders vs current climbing I’m on doing on an aquarium sim. Try it in your next build!

0:24
Sundar Pichai avatar
⟳ JY retweeted
Sundar Pichai@sundarpichai·Nov 14

Congrats @SpaceX for that starship catch, still feels like science fiction! Watching it live was incredible. Upwards!

JY avatar
JY@agenticjay·Nov 14

My Pelican demo onstage at AI Eng is peak CT. Thank you @simonw 🙏 hope you liked it! ps the benchmark is forever yours

AI can do more of the work. So why does the job feel harder? @simonw on what happens when the easy tasks disappear and your ambitions catch up with the tools. From AI Engineer Code Summit.

1:15
JY avatar
JY@agenticjay·Nov 14

Models playing games vision only is the benchmark of my childhood

JY avatar
JY@agenticjay·Nov 14

ASK: Show us what you’re shipping! (From the guy who runs the @GoogleAIStudio gallery)

what a crazy week!! All the teams at @GoogleDeepMind poured so much into shipping Gemini 4 Carbon, Gemini Decisions API, and Gemini 4 Hydrogen. met so many people from the community already sharing the excitement & how they are enjoying the new models. we’re back heads-down to keep pushing, and we’d love two things from you: 🔥 your builds: tag us in your demos, cool builds, and examples. we’ll share favorites internally and on our socials. 🗒️ your feedback: Tell us what’s working well and what’s not. drop your links and thoughts below!

Post attachment
JY avatar
JY@agenticjay·Nov 13

This week's ships culminating in a friday afternoon demo 😅 Grew up on an IRL version of this one, fun nostalgic build with cozy vibes What's interesting is the first level or two I can always win, but after that the model is able to reason ahead of me across multiple moves. While it thinks longer, when it figures it out it flys thru all the right moves where a human clicks and undoes their quick moves multiple times to test and play with solutions.

JY avatar
JY@agenticjay·Nov 13

Gemini 4 Carbon, Hydrogen, and the Decisions API in one game: Parking Bench. Carbon built the tabletop puzzle in @antigravity and you race 1v1 against Hydrogen or Gemini Decision 1 on the Gemini API across 5 levels. Both models play pure vision off a single board PNG with zero grid state.

1:04
JY avatar
JY@agenticjay·Nov 11

Today we're launching the Gemini Decisions API powered by our new Gemini Decision 1 model in @GoogleAIStudio. Built for high-speed routing, policy checks, and grading where you want calibrated probabilities instead of generated text. One 150-250 ms forward pass evaluates up to 1,000 questions across text, images, audio, video, and PDFs at $0.10 per 1M input tokens with $0 output. 🤯 Try it for free in the Gemini API today!

0:13
fofr avatar
⟳ JY retweeted
fofr@fofrAI·Nov 10

A couple of early render tests with MuJoCo physics

0:06
0:07
fofr@fofrAINov 10

Carbon: - wrote a custom CUDA/C++ path tracer from scratch - integrated MuJoCo rigid-body physics - designed and tuned a 60-second Rube Goldberg machine (120,000 gravity and contact steps in MuJoCo, no keyframes) - made the 3D studio and synced a continuous camera take to a custom Lyria 3.5 track

1:00