---
title: 'NEW Gemini Update is Absolutely INSANE!'
source: 'https://youtube.com/watch?v=16557osZcMs'
video_id: '16557osZcMs'
date: 2026-11-17
duration_sec: 480
channel: 'Julian Goldie SEO'
---

# NEW Gemini Update is Absolutely INSANE!

> Source: [NEW Gemini Update is Absolutely INSANE!](https://youtube.com/watch?v=16557osZcMs)

## Summary

Google has announced Gemini 4 Carbon, its most powerful AI model yet, designed for complex, multi-step projects. The update introduces a groundbreaking 'Swarm' feature that turns a single prompt into a team of AI agents working in parallel, alongside impressive 3D and science demonstrations. While promising, the model is not yet available to everyone and requires careful verification of its outputs.

### Key Points

- **Gemini 4 Carbon Announced** [00:03] — Google announced Gemini 4 Carbon on November 10th, describing it as their strongest AI model yet, built for complex tasks and multi-step planning.
- **Carbon vs. Argon** [00:28] — Argon is for everyday tasks, while Carbon is for harder, more complex jobs. Carbon thinks deeper and takes longer, so it's not for simple queries.
- **Swarm: AI Agents Team Up** [01:31] — Swarm, a preview feature in Google Anti-Gravity, allows a single prompt to create a team of AI agents that split up work, work in parallel, and check each other's output.
- **Google's 3D Demonstrations** [03:13] — Google showcased four examples: an interactive 3D model of the Eiffel Tower's construction, an open-world game as a pelican on a bike, a textile visualizer from a research paper, and a 3D recreation of an office from planning PDFs.
- **Science and Security Applications** [04:07] — Carbon was used to help prove code for the Signal app, find hidden protein relatives (40x more on average), and spot small problems in lab experiments.
- **Benchmark Results** [04:47] — Google claims Carbon ranked #1 on Cursor Bench 4.0, Perplexity Wonder, and Zapier Automation Bench, and #2 on the multimodal knowledge agent test.
- **Security Improvements** [05:13] — Carbon blocked 99% of attack paths, with a 0.8% break-in rate per try (vs. 3.1% for Argon), and scored 92% on an honesty test for agents.
- **Availability** [05:56] — Carbon is available to developers via API, businesses via Antigravity, and rolling out to Google AI Ultra subscribers. It is not on the free plan.

## Transcript

Today I'm going to show you the new Gemini update is new, it's big, and it's kind of wild. Google dropped Gemini 4 Carbon, and I'm going to break it down in plain words. First, what it is, then the swarm. That part blew my mind. And
later, I'll show you prompts you can copy. Let's start simple. On November 10th, Google announced Gemini 4 Carbon. Google says it's their strongest AI model yet, but it's not just a faster chatbot. It's built for hard stuff, big
plans, it tries different ideas, then it double checks everything. Now you might ask, what about Gemini 4 Argon? Good hard work. Google says Argon is still
tasks. Carbon is what you bring in when the strong player. Carbon is the closer. Carbon also thinks more than Argon. It digs deeper. That can take longer, so don't use it for tiny stuff. But one
feature in this update stole the show. I'll get to it soon. First, let's make this real for your business. Say you run an online community like my AI Profit Boardroom, you could give Carbon a giant task, something like this. Plan a full
launch for the AI Profit Boardroom, design a landing page that converts, show the value of AI automation, explain every benefit of our community, help us get more customers, leads and traffic, write the emails, list the risks, then
not one little question. That's a whole project, and that's what Carbon is built for. Now it's not perfect. Google says Carbon can still make things up and it can time out sometimes. So always check the work. But here's what's cool. Most
AI answers and forgets. Carbon plans, tries, and verifies. Here's the part that got me. Google is previewing something called swarm. It lives inside Google Anti-Gravity. You type slash swarm. Normally you ask one AI agent to
do everything. That's like asking one person to build a house alone. Slow, messy. With swarm, one prompt turns into a whole team of AI agents. Here's how it works. You give it a big project. The agents split up the work. Each one gets
a job. They all work at the same time in their own separate workspaces, so nobody steps on anyone's Carbon can take the hardest jobs, like planning the build, running the team, and checking the final work. Then you have checker agents. They
problems. Think of the grumpy teacher with the red pen. And you can watch the whole team working in real time. One note, Swarm is a preview, so it's early, but the idea is huge. We're going from AI that writes code to AI that runs a
whole work process. And wait until you hear what Google built with it. Quick If you want to learn how to save time and automate your business with AI tools like Google Gemini, check out my AI Profit Boardroom. It's the best place to
scale your business, get more customers, and save time with AI automation. Here's what you get. We've got a 30-day Gemini roadmap. Day one, you learn what Gemini for Carbon can do. Day two, you write your first big prompt. Day three, you
business. And it keeps going from there, one simple step at a time. You also get four coaching calls every week. So if you get stuck, you can ask and get help. AI tools like it inside. So you're not learning alone. You're learning with
you've been watching AI updates and thinking, how do I actually use this in my business? That's exactly what the AI Profit Boardroom is for. Link in the aiprofitboardroom.com. Google shared four examples. Let's go fast. Number
one, the Eiffel Tower. A team of Carbon agents took old photos and engineering drawings. They turned them into an interactive 3D show of how the tower was interactive 3D show of how the tower was built between 1887 and 1889. Number two,
an open world 3D game. You explore Google's Mountain View campus as a pelican on a bike. I'm not joking. The agents handled the 3D art, the code, the three, a science tool. Carbon built an interactive textile visualizer from a
research paper. It had 3D yarn, fiber tracking, and live stress curves. Number four, an office rebuilt in 3D. The team used seven public planning PDFs with 206 drawing views, over 10 overnight loops. They built Google's Platform 37 office
in 3D, about 906,000 triangles. These aren't just chat answers. These are full projects. I haven't tested these myself. These are Google's own examples. I to be honest with you. But it's not just about code. Google also points to science.
First, alpha proof code. Researchers used carbon with other models. They turned security goals into math rules. Then they wrote millions of lines of Signal is the secure messaging app. Quick note. That does not mean the whole
signal app is proven safe. Let's keep it real. Second, protein research. Researchers used carbon in a team setup to find hidden protein relatives in huge data sets on seven protein families from a test called CASP17.
Google says they found about 40 times more on average. Third, lab work. Google says carbon helped spot small problems in experiments like mixing that wasn't finished and tiny bits of leftover samples. So carbon isn't just a coding
researchers. Now let's talk results. Google says carbon ranked number one on three tests, Cursor Bench 4.0, Perplexity Wonder, and Zapier Automation Bench. On a test called multimodal knowledge agent, it ranked number two.
That test checks how AI handles charts, tables, drawings, and long documents, real work stuff. But be careful. These are Google's own results. They don't prove carbon beats every model at everything. Treat it as a good sign, not
because this matters. Carbon can help defenders find weak spots in software and build and check fixes. Google shared some test numbers. 99% of attack paths were blocked. The break-in rate per try was 0.8%. On Argon, it was 3.1%. On an
honesty test for agents, carbon scored 92%. Argon scored 70 %. And on one prompt injection test, attacks worked 0% of the time. But hold on. Those are test results, not a promise of perfect safety. Google itself says prompt
that part is not open to everyone. So who can use it? Developers can use it through the Gemini API in AI Studio and Android Studio. Businesses can get it Antigravity, and Code Mender. Scientists can apply to a research program. For
everyday users, it's rolling out to Google AI Ultra subscribers in the Gemini app, in Antigravity, and in AI mode in Google Search. It is not on the free Gemini plan. So don't expect it tomorrow if you're on free. One more
thing, it handles up to 1 million tokens and it takes in text, images, audio and video. You can also set how hard it thinks. More thinking can mean slower answers. So here's my take. Gemini 4 Carbon is a big statement from Google.
We're moving from AI that answers questions to AI that plans, builds, checks and works in teams. Swarm is the star. The 3D demos are the eye candy and the science work makes it feel serious, but be smart. Swarm is a preview. The
for everyone yet. So tell me, is this the most exciting AI release this year? And what would you build with a team of AI agents? Comment below. Julian Goldie reads every comment. Now, before you go, there's a window right now where most
business owners haven't figured this out. They haven't learned how to use AI everything by hand. That window won't stay open forever. Soon everyone will be learn it is right now. If you want to get ahead of the crowd, join my AI
Profit Boardroom. Here's what's inside. Daily tutorials so you can learn tools like Gemini step by step. Four coaching calls every week so you can get your with prompts like the ones I showed you today for landing pages, emails and
launch plans and a member map. All of it is built to help you save time, get more customers and automate your business with AI. So if you want to use tools like Gemini 4 Carbon the right way, this is the place to start. Link in the
aiprofitboardroom.com. If you want the full process, the SOPs and over 100 AI use cases like this one, join the AI Success Lab. It's our free description. You'll get all the video notes from this episode. That means the
ideas and steps I covered today all in one place. Plus you get access to our community who are crushing it with AI. So if you want to keep up with Gemini it's free. Links in the comments and description. Thanks for watching. I'll
description. Thanks for watching. I'll see you in the next one.
