Claude + Antigravity + Stitch + Nvidia (AI NEWS)

Claude and the Gravity Stitch in NVIDIA. So this week Google just dropped four things that most people haven't even heard of yet. And if you're a business owner, a freelancer, or anyone who does any kind of work on a screen, these tools are going to change how you work. Maybe this year, maybe this month, I'm not being dramatic.

So Google just upgraded a tool called Stitch, turning it into what they're calling an AI native design canvas. Anyone can now describe what they want in plain English and Stitch builds it. A full app design, a website, a product screen from words. Google AI Studio just got a full upgrade that lets you turn prompts into real working apps with multiplayer features, databases, real logins, and modern web tools.

And they built it on top of the Google anti-gravity coding agent. And that's not even the wildest part of this week. NVIDIA announced something called NemoClaw, a security stack for OpenClaw that lets you run powerful AI agents locally on your machine privately for free. They launched it GTC 2026.

Jensen Huang, the CEO of NVIDIA, one of the most powerful companies on the planet right now, stood on stage and actually said that every company in the world needs an OpenClaw strategy, every single company. And whilst all that good stuff is happening, there's a dark cloud over at OpenAI. Internal projections show OpenAI is headed for $14 billion in losses in 2026 alone against roughly $13 billion in sales. That means they spend more than they make by a billion dollars with the largest AI product on earth.

We'll get into what that means in a minute, but first let's talk about what Google just did, because it matters to you right now, today, whether or not you've ever written a single line of code in your life. So let me paint you a picture of what used to happen when someone wanted to build an app, right? They had an idea. Maybe it was a booking system for their small business.

Maybe it was a tool to manage their team schedule. Maybe it was a landing page that looked clean and modern. They'd have that idea and they'd hit a wall. Because to build something on a computer, a real thing, something people could actually use, you needed a designer and then a developer, and then someone who knew how to connect the design to the code, and then someone who knew how to put it on the internet, and then someone else who knew how to add a login screen, a database, and a payment system.

That was a team. That was months. That was tens of thousands of dollars. And most small businesses just give up, right?

They used a basic template on Squarespace or Wix and just called it a day. And there's nothing wrong with that, but it meant the gap between what a big company can build and what a small business owner can build was enormous. And that gap is closing fast. So let's start with Google Stitch.

Stitch is a tool built by Google that uses AI to translate interface designs into HTML and CSS code. And the tool also supports Tailwind, a widely used development framework. Here's what that means in plain English. You want to build a website.

You have to have it as a signup page, a dashboard, a checkout screen. You type that in like a text message. You say, I want a checkout page for a software product, clean design, green and white colors, modern feel. And Stitch goes off and actually builds it.

Now, here's the thing. When you're using this, right, a finished screen with real code behind it that developer can actually use comes up. And the upgraded version can now generate up to five screens at once. The old version did one at a time, but it's the thing that made Figma's stock drop 8% in just two days.

Google redesigned Stitch with an infinite canvas, voice interaction and connections to coding assistants, including Cloud Code and Cursor. And the market impact was immediate with Figma shares dropping 8%. Now, Figma is a company worth billions of dollars. It's the tool that professional designers use all over the world.

And Google just shipped a free competitor that does a first draft in 20 minutes from a text prompt. Google Stitch gives you 353 design generations per month, exports to Figma format, exports to React code. All you need is a Gmail account. Now, let me say that again, free Gmail account, 350 designs per month.

You can do your entire business's website, your entire app, every screen, every page for free. Now, I want to be real with you here because Stitch doesn't do everything. It doesn't capture the deepest strategy behind design, doesn't replace a great designer who understands your customer deeply. If you're running a massive brand with precise design language, you still want humans involved.

But for a first draft, for showing a client what something could look like, for building a prototype in a day instead of a month, this changes everything and it gets better because Stitch doesn't actually work alone. And this is where it starts to feel like science fiction. There's a workflow that's become one of the most powerful combinations in the builder space right now. You use Stitch to create a polished visual design, then pass it to Antigravity to develop it into a fully functional application.

Stitch designs it, Antigravity builds it, two tools, both free with no code required. Google Antigravity is a development platform that combines a familiar coding experience with a new agent-first interface. It lets you deploy agents that autonomously plan, execute and verify complex tasks across your editor, terminal and browser. So think of it this way, you know how you can hire someone on Fiverr to build a website, you tell them what they want, they go away, they come back with something done, you review it, give feedback and fix it.

Well, Antigravity is like that, except it runs on your computer, it costs nothing and it works in minutes instead of weeks. Antigravity also includes Claude Opus 4.6, one of the strongest models for agentic coding and no charge. So running Claude code on complex code-based tasks through the API often costs like $5 to $15 per session. Teams doing daily development can hit $100 to $300 a month in API costs.

Antigravity removes that cost entirely. Now, let me be really clear about what that means, right? Claude Opus is one of the most powerful versions of Claude that Anthropic makes, the one developers pay serious money to access through the API, the one big companies use, and Google is giving it to you for free inside Antigravity because Google wants you using Google Cloud, that's a trade. You get the tool for free, they hope you deploy on their infrastructure, and if you don't, you still get a free tool that builds real apps.

For creating new applications, the ROI is massive. You got a developer can spin like a full-stack prototype in hours rather than days, which accelerates time to market and allows for faster validation of business ideas. So hours instead of days, free instead of thousands of dollars. And then Google did one more thing this week.

Google AI Studio launched a completely upgraded experience designed to turn your prompts into production-ready applications. You can now build multiplayer experiences, install external libraries, add secure storage and user authentication via a built-in Firebase integration. The translation is you can build an app that has real user accounts, real logins, real saved data, multiple people using at the same time that used to require a full engineering team. This new experience in Google AI Studio has already been used to build hundreds of thousands of apps.

Google is working on integrations with Workspace to connect drive and sheets, plus the ability to take your app from Google AI Studio to Antigravity with a single button click. One click, AI Studio to Antigravity, prompt to production. We're watching the entire cost structure of software development collapse in real time. And if you're a business owner sitting on an idea, something you've wanted to build for years but always said can't afford it, I don't know how, I need you to hear me right now.

The excuse is gone, the tools are free and the barrier is your idea not your budget. So now let's talk about NVIDIA because what happened at GTC 2026 this week is one of the biggest things in this space and most people completely missed it. You've probably heard of OpenCore. I've talked a lot about it on this channel.

You know it's the AI coding agent that runs locally on your machine, meaning it works on your files, your data, your computer, not some cloud server somewhere. Now OpenCore is one of those tools where once you understand exactly what it does, you can't unsee it. Jensen Huang actually stood on stage and said every single company in the world needs to have an OpenCore strategy. That's not a product manager, that's the CEO of NVIDIA, a company that just reported quarterly revenue of 39 billion dollars.

The man who has been more right about AI than almost anyone on the planet. Every company, OpenCore strategy. But here's the problem that's been holding OpenCore back from being fully trusted by businesses. As AI moves from simple prompts to agentic AI, long-running systems, reason, plan and act, developers need secure infrastructure to build always-on assistance.

So when you give an AI agent access to your computer, your files, the apps, you're trusting it a lot and for a small personal project that's fine. But for a company with sensitive customer data, maybe a law firm, a hospital, a financial services business, well you need walls, you need rules, you need to know that AI can only touch what you say it can touch. And that's exactly what NemoCore solves. So essentially NemoCore is an NVIDIA agent toolkit that installs OpenShell to provide open models and an isolated sandbox that adds data privacy and security to autonomous agents.

It provides a missing infrastructure layer beneath AI agents to give them the access they need to be productive, whilst enforcing policy-based security network and privacy guardrails. Think of it like this, you have a really capable AI assistant who can do almost anything. But you work in a building with sensitive files, you can't just give them the keys to every room. You give them a key to the room they need and a log of everywhere they go.

And then NemoCore is a key management system for your AI agent. The first features available in NemoCore are NVIDIA Nemetron open models and the NVIDIA OpenShell runtime. Nemetron local models enable users to run inference locally, which means better privacy and no token cost. No token cost is huge, so that every time you use an AR model through the cloud, whether that's ChatGPT, Claude or Gemini, you're paying per token.

Per word, basically, if your agent is doing complex tasks all day, that can add up fast, right? Running it locally makes you pay once for the hardware and then you never pay per word again. And now this week there's another piece of that puzzle. So Nemetron Cascade 2 is now available to run through Olama.

It's a 30 billion parameter model that delivers strong reasoning and agentic capabilities on par with models that have been up to 20 times more in terms of parameters. It achieves gold medal performance in both the 2025 International Mathematical Olympiad and the International Olympiad in Informatics. And I want to sit with that for a second. Gold medal performance in the International Math Olympiad and the International Computer Science Olympiad.

These are competitions where the top human experts in the world compete. High school and college students who are the best of the best at maths and programming. And this model that you can run for free on your own computer performed at gold medal level. And you can run it locally right now for free.

All you do is you set up Olama and then you would run Nemetron Cascade 2. Now you do need a good setup to do that, but at the same time these models are becoming cheaper, smaller, free, and you can run locally, which is absolutely wild. You can also run this inside OpenCLR with Olama. So if you run the Olama launch OpenCLR command inside your terminal once you've got Olama installed and you've got Nemetron Cascade installed, you're good to go.

And that's basically it. It's like two lines to install it and also to get it working with Olama directly. Now let me tell you why this matters for regular people, not just developers. I talk to a lot of business owners and one of their biggest fears that I hear is, you know, that I don't want my business data in the cloud.

I don't want my client files going through some AI server I don't control. And that's a real concern. It's been one of the main reasons big companies have been slow to adopt AI tools for sensitive work. NemoClaw and Nemetron Cascade 2 running locally changes that conversation.

Your data stays on your machine, the model runs on your machine, nothing leaves your building unless you say so. That's a completely different risk profile than sending your client contracts to a cloud AI service. Think about what this unlocks. A small law firm can have an AI assistant that reads case files, drafts, documents, and keeps a record of every action without ever sending a single client document to a third-party server.

A medical practice can have an AI that helps with notes and records locally, privately, on-site. And a financial advisor can have an AI that analyzes portfolios and generates reports without exposing client data to anyone. This isn't science fiction, this is a combination of tools that you can set up right now today. And then there's Claude Cowork because Anthropic had a big update this week too.

Anthropic leased a persistent agent thread for pro and max plan users to manage cowork tasks from mobile and desktop. Users on pro and max plans can access this via Claude Desktop or the Claude mobile app for iOS and Android. Now here's why that's a big deal. Cowork is Claude's desktop agent.

It can access your files, your apps, your connectors. It can run multi-step tasks whilst you step away. It's like having an assistant on your computer, but until now you have to be at your desk to kick off tasks. Dispatch, the feature that lets you message your cowork session from your phone whilst your computer does the work, creates one persistent conversation between the Claude mobile app and your desktop.

Your Mac handles everything, file access, connectors, sandbox implementation. Your phone is just the messaging interface. You text Claude from your phone and your computer does the work. You come back to finish results afterwards.

This is what an AI assistant is supposed to feel like. And there's more. With the introduction of projects in Cowork, you can organize related tasks into persistent self-contained workspaces with your own files, links, instructions, and memory, making Cowork more powerful for recurring or long-running work. Projects in Cowork.

That means your AI assistant now remembers the context of your ongoing work. It knows what project you're working on. It knows your files. It knows your preferences for that project.

And it's not just a chatbot anymore. It's persistent, working as a presence on your computer that keeps getting smarter about your specific work. Anthropic says Claude Cowork goes beyond the basic chatbot interface and that real-time context from internal connectors should help employees feel like they're getting a real productivity boost. Engineers think about Claude Code as a tool they just couldn't live without anymore.

We expect that every knowledge worker will feel that way about Cowork. That's what Claude has said. Every knowledge worker. That's a bet.

That's where this is going. Now, let me stop and be honest with you for a second. Because I know some of you are watching this and thinking, okay Julian, you say this stuff every week. Every week there's a new tool that's going to change everything.

And I'm still sitting here not doing anything different. I get it. I really do. There's something called AI fatigue.

When you hear about so many tools, so many updates, so many announcements, that it all kind of blurs together into noise. And you stop taking any of it seriously because it's just too much. And I hear that. I understand that.

But here's what I'd ask you to notice. A year ago, these tools didn't exist. 18 months ago, building an app without code was a party trick. It was a toy.

Something you show your friends and laugh about. Something that you wouldn't use to run a business. But today, Antigravity scores 76.2% on SWE Bench Verified. A benchmark testing AI agent's ability to resolve real GitHub issues.

For context, Devin AI scored 13.86% when it was launched in 2024. So that's 76% versus 13% in less than two years. That's not a trend line. That's things changing at an exponential rate.

And the people who learn how to use these tools, not perfectly, not as experts, but just well enough to apply them to their actual work, those people are going to have a massive head start on everyone who waits. That's why I actually built the AI Profit Boarding. When I started learning this stuff, I kept running into the same problem. There were tutorials everywhere, YouTube videos, blog posts, Twitter threads, but none of it was designed for the kind of person I was trying to help.

For example, a business owner, a creator, a freelancer. Someone who doesn't want to become a developer. Someone who just wants to know, what do I actually use and how do I use it in my business right now? The AI Profit Boarding is that.

Every week, you get four live coaching calls on AI automation, real workflows, real tools, real applications for real businesses, daily tutorials. You get step-by-step video guides, so you can actually follow around and do the thing rather than just watch someone else do it. And you also get 30 day roadmaps. You have a clear path from where you are now to where you want to be.

And it's a community of 2,600 business owners who are already using these tools. People who have already figured out the workflows, who've already made the mistakes, so you don't have to. And something that I'm really proud of is a local map feature. So you can actually connect, meet up and get support from people who use these tools near you.

Because sometimes, you just need to sit with someone who gets it. And there's always someone online day or night, so you'll never be stuck alone, wondering what to do next. So if you want to actually do something with everything that I talk about on this channel, and not just watch it, but actually use it, the AI Profit Boarding is the place to do that. The link is in the comments inscription, or you can go to the AIProfitBoarding.com.

Now, let's talk about the dark cloud. Because whilst Google and Nvidia and Anthropic are all shipping incredible tools, there's something at OpenAI that you need to know about. OpenAI could burn through the last of its cash by mid-2027. That's not a random number I made up.

That's from analyst projections published in major financial press. The company projects $14 billion in losses for 2026 alone against roughly $13 billion in sales, with total spending clocking in at around $22 billion. They make $13 billion, they spend $22 billion. Now, before you panic, before you start thinking, should I stop using chat GPT?

Let me give you the full picture. OpenAI has raised an enormous amount of money. OpenAI has raised more than $60 billion since 2015, including $41 billion last year in a round led by SoftBank, that was the largest private fundraise ever seen. They are not shutting down tomorrow.

But here's what this actually tells us about the AI industry. The company expects cumulative losses of $115 billion through 2029 before reaching profitability sometime in the 2030s. Think about that. Cumulative losses of $115 billion.

That's not a tiebreaker. For context, the entire GDP of Bulgaria is about $100 billion. OpenAI plans to lose more than that before it makes money. Now, they have a plan.

Their own internal forecast project over $100 billion in annual revenue by 2029, which is roughly half from chat GPT and about 20% from API sales to developers. Sam Altman has also said revenue could surge to $100 billion by 2027. And that's an extremely ambitious claim. And here's the thing.

I'm not saying OpenAI is going to fail. I'm not saying chat GPT is disappearing. They have Microsoft behind them. They have SoftBank.

They have some of the most powerful investors in the world. But I am saying the economics of building AI are absolutely brutal right now. OpenAI is carrying roughly $100 billion in debt. And experts warn investors continue to pour money in because of fear of being left out, not because the numbers make sense yet.

And that's what economists are calling it, right? Fear of being left out. That was the real implications of what they're meaning. Now, here's what this means for you.

The big companies are spending hundreds of billions of dollars to build these tools and losing money, which means right now you're getting access to some of the most powerful technology built at prices that are often free or near free because these competitors are in a race to win your loyalty before anyone else does. This is not going to last forever. The window is open. It won't always be.

All right. So I would make the most of it whilst you can. Now, let me connect all of this together for you because there's a thread running through every story I just told you, right? Google builds Stitch and Gravity, anti-gravity free for anyone with a Gmail account.

NVIDIA ships NemoClaw and Nemotron Cascade 2 for free, open source, run it at home. Anthropic upgrades, co-work with projects and mobile access for $20 a month, and OpenAI burns $14 billion, building the most powerful chatbot in history, available to anyone for free. Every single one of these companies is losing money on you right now because they want to build your habits, your workflows, your whole business around their tools. So when the prices go up, and they will, you're too dependent to leave.

This is the land grab. This is a moment before the fences go up and you have two choices. You can watch from the sidelines, wait until these tools are more mature, wait until there are more tutorials, wait until it feels less overwhelming, or you can start right now whilst everything is free, whilst the community is small enough that the people who figure this out will be years ahead of the people who will figure it out later. So this is the thing really, you really want to focus on like, okay, what can you look at right now and what does this mean for you?

I mean, let me give you the scorecard for this week in simple terms. Google is shipping loads of stuff, NVIDIA is shipping loads of stuff, Anthropic is also creating a lot of stuff, and OpenAI is burning through cash just to keep you loyal. And it's a signal that the economics of AI are actually harder than the hype suggests. So here's what I want you to do with this video in terms of actionable takes.

If you've never tried antigravity, go to antigravity.google, sign in with your Gmail, and describe a single thing that you've wanted to build for your business. Watch it happen. Don't aim for perfect, but just see what it does. And if you're using OpenIV or thinking about it, go look at Nemetron Cascade 2 on Olama.

Run it locally if you can, I'd recommend having a Mac Studio for that. Even on a reasonably modern laptop, some of these models can work quite well, but the idea that you need to run a supercomputer to run AI locally is already outdated. Now, if you're already a cloud user, you can download cloud desktop if you haven't already. Turn on co-work, set up one project for your most active ongoing work, and just give it your files and let it surprise you.

And if you're a business owner who's been watching all of this from the sidelines thinking, I need to figure this out, but I don't really know where to start, the AI Profit Boardroom is exactly where you start. Real coaching calls, real tutorials, real community of people who are already doing this stuff in real businesses. Not theory, not hype, the actual workflows that work. Now, the thing that I keep coming back to this week is something Jensen Huang said, every company needs an open law strategy.

He didn't say every tech company, he didn't say every startup, he said every company. That means the bakery on your street, the accountant you use every April, the therapist, the contractor, the salon, the person who runs a food truck, the person who teaches piano lessons, the person who sells handmade jewelry on Etsy. Every single company. Now, obviously a bakery doesn't need super complex AI agents running on NVIDIA hardware, but the principle is right.

Every business at some level is going to need an AI layer. It's going to automate something, it's going to speed something up, it's going to lower costs somewhere. And the businesses that figure this out first will have margins, their competitors can't match, they'll move faster, they'll do more of less. And the businesses that wait, not because they made a considered decision, but just because they never got around to it, will wake up one day and just wonder why are they falling behind.

I don't want that to be you, okay. I've been saying this for months, the tools keep getting better, the cost keeps going down, and the barrier keeps getting lower as well. And I know some of you feel like you've heard this before, like AI is always almost there, like it's always this is the week it gets real. But look at the actual numbers.

This stuff is changing fast. Antigravity scoring 76.2% on automated coding benchmarks. You know, you've got tools like Devin that in comparison were scoring 14% before. You've got Nemetron Cascade scoring gold medal performances in both the International Math Olympiad and the International Olympiad in Informatics.

And you've got Anthropx valuation that's grown to $380 billion. Plus that company says every knowledge worker will be using something like Co-Work in the future. So these are not vibes, these aren't benchmarks, these are valuations, capability scores. The curve is steep and we're still really early right now.

So the question that keeps me up at night right now isn't will AI change business? It already is. The question is who is ready when the change accelerates? Because here's the honest truth that nobody likes to say out loud.

The people who are learning these tools right now, who are building the habits, who are making the mistakes and figuring out the workflows, they are not going to be equally affected by what's coming as the people who waited. The person who's been using AI to automate the boring parts of their job for six months has a completely different relationship to AI than the person who downloaded ChatGPT once, asked it three questions and then just decided it wasn't that useful. One of those people walks into the next wave of AI advancements with skills and context and intuition about what works and what doesn't. The other starts from scratch.

I keep saying this and I'll keep saying it because I really mean it. The time to learn is not when everyone else is learning. That's when it becomes table stakes. The time to learn is right now when most people are on the fence.

The tools are free, the resources are available, the community exists. What's the reason to wait? So that's it from me today. That's it for the update this week.

It's been an amazing week. Lots of cool stuff going on and if you haven't already, make sure you subscribe. Drop me a comment below. I would love to know what you think and I'll see you on the next one.

Cheers, bye-bye.