Nemoclaw + Minimax 2.7 + OpenClaw + GPT 5.4 Mini + Claude Dispatch (AI NEWS)

Nemo Claw, Minimax 2.7, Open Claw, GPD 5.4, Mini, Claw Dispatched, all released in a week. This week in AI, I need you to pay attention, because what dropped between March 16th and March 19th, 2026, is not a normal news cycle. This is not the usual, here's a new model, here's a benchmark, here's a percentage improvement kind of week. This week, an AI helped build itself.

A Chinese lab released a model that wrote its own code and fixed some bugs during its own training. A model that handled up to 50% of its own development workflow without being told to. This week, NVIDIA's CEO, Jensen Huang, stood on a stage in San Jose and told every CEO in the room, directly by name, that they need an Open Claw strategy. He compared it to Linux, he compared it to the invention of HTTP, the protocol that built the internet.

This week, Anthropic shipped a feature that lets you send your AI to work whilst you're at the gym, from your phone, whilst your laptop sits on your desk at home, just doing the work. This week, Google opened up a feature that reads your Gmail, your Google Photos, your travel bookings, your purchase history, and uses it all to answer your questions before you even finish asking them. And they gave it to everyone for free. This week, Google also dropped a major update to a design tool called Stitch, that lets non-designers build fully functional apps just by talking.

And Figma's stock dropped 9% the same day. And also, OpenAI released small models that are so cheap and fast, they're going to change what it means to build software. And Mistral, the European AI lab, dropped a 119 billion parameter model that runs on your own computer. All of that in one week.

I keep saying the pace of this is unlike anything I've ever seen before, I keep saying it because it's true. And I know some of you are tired of hearing of it, I understand. But being tired of hearing doesn't make it slower, it just means you're less prepared for what's coming. So let's go through all of it story by story, what happened, why it matters, and what you should actually do about it.

Let's start with the one that hit me the hardest. Minimax M2.7. Released on March 18th, 2026, Minimax is a Chinese AI company, they've been putting out models in the M2 series at a pace that's been genuinely impressive. M2 last year, M2.1, M2.5 in February, each one faster, cheaper, and more capable than the last.

But M2.7 is different. And the reason it's different isn't the benchmarks. The benchmarks are great, but they're not the full story. The story is what the model did whilst it was being built.

When Minimax was training M2.7, they actually set up a system where earlier versions of this model could participate in their own training process. They called it an agent harness. They were it like this. They gave the AI a workspace, tools, access to logs, the ability to run experiments, and they said, go figure out what's wrong and fix it.

And the model did. It was reading its own error logs, running tests, spotting patterns in its own failures, trying different approaches, running up to 100 rounds of iteration on a single problem, and then applying what it learned to improve itself. The result was a 30% improvement on their own internal tests, just from the model working on itself. And here's a number that really matters.

Minimax says M2.7 handled between 30 and 50% of its own development workflow. That means if you had a 10 person research team building this model, somewhere between three and five of those jobs were done by the model itself. I want you to think about what that means for the future, not in some abstract sci-fi way. In a practical way, we've always been assuming improving AI requires humans.

Brilliant, expensive, hard to hire humans. That's been the bottleneck. The thing that was supposed to slow this down, human researchers can only work so fast. There's only so many of them.

Training runs take time. Minimax just punched a hole in that assumption. If a model can handle 30 to 50% of its own development today, what does that number look like at six months? At what point does the feedback loop stop needing humans in it at all?

I'm not saying we're there. We're not there. But we took a real documented measurable step in that direction this week. And I think it deserves more attention than it's getting.

Now the actual performance numbers, M2.7 scored 56.2% on SWE Pro. That's the test that measures whether AI can do real software engineering tasks. Not toy problems, real world tasks, 56.2% puts it right in the conversation with the best models from OpenAI and Anthropic. And it scores 50 on the Artificial Intelligence Index.

The composite score that measures reasoning, maths, knowledge, and coding combined. The medium for models at its price range is 19. It absolutely crushed models in its price tier. And the price itself, $0.30 per million input tokens, $1.20 per million output tokens.

For a model at this level of performance, that's a very low number. There's also something worth noting about the direction Minimax is going. They used to be open source. They released M2 with open weights.

Anyone could download it, run it, use it for free. Developers love them for it. But M2.7 is proprietary. You access it through the API or the agent platform.

The weights are not publicly available. That shift is a signal. It means Minimax believes this model is good enough to charge for. And looking at the benchmarks, they're probably right.

And they're not alone. Z.AI went proprietary with GLM5 Turbo. There are reports that Alibaba's Quen team is shifting direction to the era of Chinese AI labs, giving away frontier models for free, maybe ending. When your model is competitive with GPT and Quad, you stop giving it away.

But here's the thing about M2.7 that I keep coming back to. The evolution story isn't just about one model. It's a proof of concept. It's a demonstration that loop can work, that AI can participate in its own improvement in a meaningful, measurable way.

Every lab that reads that result is going to try to replicate it and improve on it. And the models that come out of that process are going to be better than what we have today in ways that are harder and harder to predict. Now, let's talk about what happened at GCC. Jensen Huang took the stage in San Jose on Monday, March 16th, in front of a packed arena.

GCC is NVIDIA's developer conference, and this year it ran through March 19th. It's actually still happening today as we record this. The headline announcement was something called NemoCore. But before I explain what NemoCore is, I need to make sure you understand what OpenCore is because if you don't understand OpenCore, NemoCore makes no sense.

And OpenCore is something you should absolutely know about at this point. OpenCore started as a side project from a developer named Peter Steinberg. It built a framework, basically a set of tools that let you create AI agents that run in your computer, not in the cloud, on your machine, using your files, your apps, and your data, and it exploded. 320,000 GitHub stars in two months.

By early March, it had become the most starred project on all of GitHub, surpassing React, Linux, the biggest open source projects in the world, by that measure, in two months. It happened because the combination of cheap open source AI models and a good framework for running them locally was exactly what the developer community had been waiting for. DeepSeek made powerful models for free, OpenCore made them useful, and suddenly anyone with a computer could have an AI agent that worked with their actual files, connected to their apps, and could be controlled from a messaging app like a Telegram or WhatsApp. Jason and I actually acquired OpenCore in February.

They saw what was happening and moved fast, but OpenCore had problems, right? Real problems. Security vulnerabilities, prompt injection attacks, cases where agents didn't do what they were told to. Meta reportedly banned OpenCore from work devices after a documented incident where an agent accessed an employee's machine without instruction and deleted her emails in bulk without asking.

So, the dream was real, the safety wasn't. That is a gap NVIDIA is walking into. Jensen Huang on stage said, Mac and Windows are the operating systems of the personal computer. OpenCore is the operating system for personal AI.

This is the moment the industry has been waiting for. He didn't call it a useful tool, he called it an operating system, the foundation layer of a new era of computing. He compared it to Linux, he compared it to HTTP, the protocol that built the entire world web. And then he said this, for the CEOs, the question is, what is your OpenCore strategy?

We need it. We all have a Linux strategy, we all needed a HTTP strategy which started the internet. What is your OpenCore strategy? Not someday, now.

So what is NemoCore? Well, in plain English, it's OpenCore with enterprise-grade security and privacy built in. So you install it with one command, just one command, and you have a fully functioning, fully secure AI agent environment running on your own hardware. It installs something called OpenShell, NVIDIA's new runtime for running agents safely.

It creates a walled sandbox so the agent can only see and touch what you explicitly give it permissions to see and touch. It has a privacy router that decides which AI goes to the local models on your machine and which one goes to the cloud models, and you define the rules. It's hardware agnostic, so it doesn't need NVIDIA chips, it runs on anything, but if you have NVIDIA hardware, the new DGX Spark and RTX Pro workstation, you can run their Nemetron models locally with no internet, no cloud, no monthly API pill, and no data leaving your machine. The model they're pushing for agent work is Nemetron 3 Super, which is 120 billion parameters.

On a new benchmark called PinchBench, which specifically tests how models perform inside OpenClaw, Nemetron 3 Super scored 85.6%, top open weight models for agents in its class. They also released Nemetron 3 Nano 4B, that's a tiny version, for regular consumer GPUs, a normal gaming PC with an RTX card. If you have that, you can now run AI agents locally on it. No subscription, no API key, just your computer, your data, your agent all the time.

I want to be direct about why this matters for businesses. The companies that build AI agent workflows into their operations right now, that automate the repetitive, the predictable, the time consuming, are going to have a cost structure that their competitors can't match. Not eventually, not in theory, in the next 12 to 24 months, a business that has automated its research, its first drafts, its data pulling, the scheduling, its client communication handling, that business can run leaner. It can respond faster, it can do more with fewer people, and it creates a gap that compounds.

Jensen Huang's message to CEOs was not think about this, it was, you really needed a strategy, you're behind if you don't have one. Now, Claude Dispatch. This one launched on March 17th and it's genuinely exciting, even though it's still early and has real limitations. I want to be honest about both.

Anthropic released Dispatch as a research preview. It's a future, a feature, inside Claude Cowork, their desktop AI agent. And here's what it does in plain terms. You leave your laptop at home, you go somewhere, you pull out your phone, you open the Claude app, you type, take the sales data from my spreadsheet, compare it to the last quarter, write a brief report, and save it to my Google Drive.

You put your phone away, you go about your day, and when you come back, the work is done. That is Dispatch. One persistent conversation thread that syncs between your phone and your desktop. Claude runs on your computer with full access to your local files, your connected apps, your email, your slack, your google drive and you direct it from anywhere.

The person who shipped this, Felix Riesberg at Anthropic, made the announcement on March 17th and the reaction in tech circles was immediate. People started comparing it favourably to OpenClaude. Ethan Mollick, who is one of the most serious researchers studying AI and knowledge work, noted it directly and the line that's been circulating all week, Anthropic built OpenClaude faster than OpenAI. Because OpenAI owns OpenClaude and still hasn't shipped a polished consumer version, Anthropic just did.

Setup is simple, you download Claude Desktop on your Mac or Windows machine, you scan a QR code to pair your phone and then the persistent cowork session is accessible from the mobile app. The privacy design is worth talking about. Everything runs locally on machine, your files don't go to Anthropic servers, your data doesn't leave your computer and that is a very different security posture from cloud-based AI agents. For businesses with real data governance concerns, and there's a lot of them, it matters.

Anthropic also built in human-in-the-loop controls. Before Claude does anything destructive, like deleting a file, sending an email, moving a folder, it sends a push notification to your phone asking you to confirm you stay in control. AI doesn't go crazy and rogue. Now, the honest stuff.

Early testing from MacStories found that complex tasks succeed about 50% of the time. That's not great. The desktop has to be on and the Claude app has to be open. No background service running whilst the laptop is closed.

There's only one conversation thread, you can't run multiple tasks in parallel. There are no notifications when tasks complete. And right now it's rolling out to max subscribers first, that's between $100 and $200 a month, with pro users at $20 a month getting access within a few days. So, yes, it's early, it's rough around the edges, Anthropic is calling it a research preview for a reason, but I want to say something about research previews.

Claude Cowork itself was a research preview three months ago. It now has a plugin marketplace, schedule recurring tasks, legal and productivity plugins, and it's integrated into Microsoft 365. Research previews from Anthropic move fast and the direction is completely clear. AI that works whilst you sleep, AI that you boss around from your phone, AI that finishes the work before you sit down back at your desk.

That is what's going on. This is where it's going. And Dispatch is version one of that future. The community framing of this is worth too.

It's worth noting, right? So Anthropic built OpenClaw faster than OpenAI. That's a meme and it's not wrong. OpenAI acquired OpenClaw in February.

They haven't shipped a clean consumer version of it yet. Anthropic just did. That competitive pressure is only going to accelerate what gets built next. Before we keep going, I want to be honest with you.

All of these stories, Claude Dispatch, Nemaclaw, Minimax, M2.7, they represent the same underlying thing. AI that does work, not AI that helps you do the work. AI that does the work itself whilst you're somewhere else, whilst you're doing something else. And that creates an opportunity, a real practical immediate opportunity for people who understand how to set these things up, what to point them at, and how to manage what comes out.

That's what the AI Profit Boardroom is about. It's a community at AIProfitBoardroom.com, link in the comments description as well, where business owners, freelancers, creators, and professionals are building real workflows with these tools right now. Not theory, not demos, actual implementations. People making actual money, saving actual time, building actual systems.

If you've been watching this channel and feeling like you understand the news but you're not sure what to do next, the AI Profit Boardroom is the next step. Link in the comments description or go to the AIProfitBoardroom.com to get access. Now back to this week. Let's talk about Google Personal Intelligence because this is a big one and it's not getting the coverage it deserves.

On March 17th, Google expanded something called Personal Intelligence to all free users in the United States. Up until that point, Personal Intelligence was only available if you were paying Google's AI Pro or AI Ultra subscriptions. AI Pro is $20 a month, AI Ultra is $250 a month. Those are paid tiers.

Most people don't have them. Now it's free for everyone with a personal account in the US with a Google account. So what is it? Personal Intelligence is a feature inside Gemini, Google's AI, that connects your AI to your Google apps, your Gmail, your Google Photos, your travel bookings, your purchase history, your search history, your calendar, all of it.

Once you turn it on, Google Gemini stops being a general purpose AI assistant and becomes an AI that knows you, not generically, specifically you. Google gives some examples of how this works. Let's say you're at a tire shop and you don't remember your car's tire size. With Personal Intelligence, Gemini can search your Gmail for the purchase receipt, find the car model, and give you the exact specs.

You don't have to look anything up. You ask, it knows. You want to plan a family vacation, for example. You ask Gemini for a travel itinerary.

Without Personal Intelligence, it gives you a generic list. With it, Gemini pulls your hotel confirmation from Gmail, looks at your past travel photos to see what your family enjoyed, and builds an itinerary around your actual preferences and actual bookings. You're shopping for a bag. With Personal Intelligence, Gemini looks at your recent purchases and suggests options that match your style, including specific details like suggesting bags with gold hardware because it noticed you just bought gold shoes.

Now here's the thing I want you to hear about this. Google is not training their AI on your Gmail data. They've been very clear about this. The feature is opt-in.

You choose which apps to connect. You turn them off at any time. Gemini reads your data to answer your questions. It doesn't train on your data to improve the model.

That's an important distinction. But still, think about what this means at scale. Not just for individual users, for the AI ecosystem as a whole. For years, AI assistants were limited by what you told them.

Every conversation started from zero. You had to re-explain who you are, what you want, and what your situation is. Every single time, that friction was real. It was one of the reasons casual users didn't get much value from AI, because the overhead of getting up to speed was too high.

Personal Intelligence removes that overhead. Your AI already knows your context, already has a background. You just ask. Now the questions get shorter, the answers get more relevant, and the amount of time AI can save you grows dramatically.

Because it's not just answering questions faster, it's answering the right questions, even when you only asked a vague one. Google is calling this shift from AI that knows things to AI that knows you. And that is not a small shift. That is a fundamental change in what AI assistants are useful for.

This feature is now free in the US. It's available in AI mode, in Google Search, the Gemini app, and in Gemini inside Chrome. It's all rolling out now. And Google Stitch.

This one moved markets. On March 18th, yesterday, Google Labs released a major update to a tool called Stitch. And within hours, Figma's stock dropped 9%. Let me tell you why.

Figma is the tool that designers use to build apps. If you've ever had a website made, if you've ever had an app designed, almost certainly the person designing it used Figma. It is the industry standard for UI design, building what things look like before a developer builds how they work. Figma is worth billions of dollars.

It has millions of professional users. It's deeply embedded in every design and product team on earth. Google Stitch is coming for all of that. Here's what the Stitch update includes.

First, an AI native canvas. Instead of a flat design surface, you're working in a dynamic environment where an AI agent is your partner. It's always present, always watching what you're doing, always ready to suggest or implement. Second, voice controls.

So you don't click, you talk. You say, move the menu to the top or apply a dark theme or make the buttons bigger. And it happens live in real time. Third, instant interactive prototypes.

Before this update, Stitch generated static images. You could see what something would look like, but you couldn't click on it. Now you hit a button and it becomes a working prototype. Click a button in design, it takes you to the next screen.

The AI calculates the navigation automatically. Fourth, and this is the one that made designers nervous, React code export. You can take any design you've built in Stitch and export it as a fully functional React application, not mock-up code, working code, a real app that you can run. Think about what that means for the pipeline between design and development.

Right now, a designer builds something in Figma, a developer looks at it, and then the developer rebuilds it in code. That handoff is one of the most expensive and time-consuming parts of building software. Misunderstandings happen. Things get built but slightly wrong.

Revisions go back and forth. Stitch just collapsed that pipeline. Design something, export the code, ship. Google is calling this concept vibe design.

You describe the feel and function you want. The AI handles the technical execution. You communicate intent. The AI produces output, and the output is not a sketch, it's a real working application.

Figma has been preparing for AI competition for a while. They have AI features of their own, but Google is coming with model sophistication, cloud infrastructure, and the full Google stack. AI studio, Gemini, the whole ecosystem. That's a different scale of competitor.

And this is a broader pattern playing out across every design and creative tool right now. The barrier to building is collapsing. Stitch is for UI design, but the same thing happened in video, in audio, in writing, in code. The gap between I have an idea and I have a finished product is getting smaller every single week.

A freelance designer who can use Stitch can now do the work of what used to be a two or three person design and development team. A startup founder with no design background can now build a polished prototype in an afternoon instead of waiting weeks for a designer. And that is the story behind the Figma stock drop. It wasn't fear that Google would steal Figma's users tomorrow.

It was the market repricing what Figma is worth in a world where free Google tools does most of what Figma does. Faster, with less training, accessible to everyone. Now let's talk models because this week also had some significant model releases that are going to matter for builders. OpenAI dropped GPT 5.4 Mini and GPT 5.4 Nano on March 17th.

These are small models, tiny, fast, cheap versions of GPT 5.4. And here's why small models matter and why people outside of software development should actually care about this. The big frontier models, GPT 5.4, Claude Opus, Gemini 3.1 Pro are already incredible, but they're also relatively expensive and relatively slow when you need to run thousands of millions of requests. If you're building an AI system that has to do a lot of small, repetitive tasks quickly, classified emails, checking data, generating short responses, routing requests, using a big expensive model for every single step is wasteful.

It's like using a Formula One racing car to drive to the supermarket. Mini and nano models are built for these scenarios. Fast, cheap, lightweight, and specifically optimized for coding and agent workflows, meaning they're designed to work well as part of large AI systems where they handle specific steps in a pipeline. The developer community, OnX, called this one of the most consequential announcements of the week for anyone building AI-powered applications because it changes the maths on what's cost-effective to build.

AI features that were borderline affordable last month become clearly affordable when the per-token cost drops dramatically. That unlocks new categories of products. And then Mistral, the French AI lab, released Mistral 4 small on March 16th. 119 total parameters, but only 6 billion added parameters at any given time.

It's what's called a mixture of experts model. It's big on paper, but efficient in practice. It unifies Mistral's previous models into one package that handles chat, coding, and agent tasks. Critically, it runs locally on NVIDIA's DGX bar and on RTX Pro workstations, which connects directly to the Nemeclaw story.

Jensen Huang is positioning NVIDIA hardware as a place where your agents live, and partners like Mistral are making sure their models run well on that hardware. The ecosystem is being built deliberately. Now, let me take a step back for a second and give you the bigger picture because when you look at everything that happened this week, there's one clear through line. Every major development this week was about the same fundamental shift.

AI is moving from something you visit to something that lives with you. Used to be, you'd open a browser tab, type something, get an answer, close the tab. That was the AI experience. Destination, a tool when you needed it.

What this week describes is something completely different. Claude Dispatch, AI that works whilst you're away from your desk. Nemeclaw and OpenClaw, AI agents running your machine all the time. Google Personal Intelligence, AI that already knows your context before you ask.

Minimax M2.7, AI that improves without being told to. The pattern is persistence, presence, agency. AI that is not waiting to be asked. AI that is just there, working, learning, available, always.

That's a different relationship with technology than anything we've had before. It's closer to having an employee than using a tool and it changes what's possible for individuals and businesses in a way that I don't think most people fully thought through yet. A solo creator with Claude Dispatch can assign research, writing, and formatting tasks from their phone on a Monday morning and come back to finish drafts. A small business owner with an OpenClaw setup can have an agent monitor their inbox, triage client requests, pull relevant data, and prepare briefs all day every day at zero marginal cost.

A consultant using Google Personal Intelligence can answer client questions with context that used to require manual digging instantly. The gap between people who know how to set this up and people who don't is not a small gap and it's growing every single week. Now let me talk about what this means for different kinds of people watching this because I think the answer is different depending on who you are. If you're a business owner the question is not whether to use AI.

That question is already answered. The question now is what specifically to automate? What are the repetitive, time-consuming, predictable things in your business that an agent could handle? Research, first drafts, scheduling, data pulling, client responses.

Pick one. Build the workflow, test it out. The cost of running these tools has dropped to the point where even a small business can afford persistent AI workflows. The ROI is not theoretical it is immediate.

If you're a freelancer or creator the competitive pressure is real and it's coming from multiple directions. Most people and more people can do what you do faster with less training because of tools like Stitch, like Claude, like GPT 5.4 and the answer is not to compete against AI and speed and volume because you're going to lose that race. The answer is to use AI to produce more, better, faster than you could alone and to focus your human energy on the judgment, relationships and the creative direction that AI cannot replicate. The tool is not your enemy, the tool is your multiplier.

If you're a designer or a developer Stitch's React export should get your attention. The handoff between design and code is where enormous amounts of time and money are lost in every production team. Tools that collapse at gap don't replace designers and developers, they change what designers and developers do. Your job becomes more about direction and judgment and less about manual execution.

That's not worse but it requires you to adapt. If you're a knowledge worker in marketing, research, finance, operations, HR, the agents in this week's news are being built specifically for you, not for coders. For people who work with information, Google personal intelligence, Claude Dispatch, Nemeclaw with business workflows. These are not developer tools, they are tools for people who manage information, write things, analyze things, communicate things.

The shift is happening in your part of the economy right now and if you're just watching all of this and not sure where to start, that's the most common position. I get it, the volume of information is overwhelming. Every week there's more, every week there are new tools, new models, new capabilities, new things you could theoretically learn. Here's what I would say, you don't need to learn everything.

You need to learn the one or two things that apply directly to what you do and then you need to actually implement them. Not watch videos about them, not bookmark articles about them, install the thing, use it, break it, fix it, learn from that. The people who are winning with AI right now, not the ones who know the most about it in theory, they're the ones who have gotten their hands dirty with the actual tools. Now let me close with something that I think is worth sitting with.

We are three and a half months into 2026 and AI has already made it hard to plan. Every week the landscape changes, every week something that was possible or impossible is now trivial, every week the cost of intelligence drops a little further and the capability goes a little higher. Minimax M2.7 helped to build itself. Think about what that sentence means for the rate of improvement going forward.

If AI is now a meaningful participant in its own development, the feedback loop just got tighter. The improvements are going to come faster, not slower. Jensen Huang compared OpenCore to the operating system for personal AI. He compared it to HTTP.

He compared it to Linux and Jensen Huang has been right about big cores before. He was right about GPUs for deep learning when almost nobody else saw it. He was right about the CUDA platform. He was right about the data center When Jensen Huang makes a comparison of that scale, it is worth taking seriously.

Google just made personal intelligence free for everyone. Gemini now knows your Gmail, your photos, your bookings, your purchases. AI went from knowing about the world to knowing about your world. That is not the same thing.

Anthropic shipped the first version of phone-controlled desktop AI. Cloud Dispatch is rough. It succeeds about half the time. But version 1 of the internet was also rough.

Version 1 of the iPhone was also rough. The direction is set. And Stitch dropped a feature that lets non-designers export working React apps from design mockups. And Figma lost 9% of its market value the same day.

This is a week. Every week has been THE week. And I know that sounds like I'm overhyping it, but I'm not making the numbers up. The AI has helped itself build itself, right?

With Minimax M217. The stock drop was real with Figma. The phone-controlled desktop agent is real. The 320,000 GitHub stars are real.

The real question is not whether this is happening. The question is what are you doing about it? And I'll say it simply. The people who are learning this stuff right now, who are building workflows, testing agents, figuring out what works, are accumulating an advantage at compounds.

Every week they learn something new. They get ahead of the people who are waiting for things to settle down. And trust me, the things are not going to settle down. So the time to start is now.

Not when you feel ready. Not when the tools are more polished. Not when you have a clearer picture of where everything is going. Right now.

Because in this particular moment in history, the people who move fast are the ones who win. That is everything from March the 16th to 19th, 2026. Claw Dispatch, Minimax M217, Nemacore and the OpenClaw ecosystem, GPT 5.4, Mini and Nano, Mistral Small 4, Google Personal Intelligence going free, Google Stitch, Drop In Voice, Instant Prototypes and React Export. Just a few days.

If you want to go deeper on any of this, if you want to actually build these workflows in your business instead of watching them just get announced, come join us at the AI Profit Boardroom. The link is in the comments description or go to the AIProfitBoardroom.com. We're building this stuff in real time and I'll see you on the next one.