AI News: The Biggest Week in AI Just Got Scary — Claude Banned, Perplexity Computer & More

AI news, the biggest week in AI just got scary. Claude was banned, Plexi computer and much more to come. So this was the biggest week in AI news that I've ever seen. This was the week the U.S.

government looked at one of the most powerful AI tools ever built. Claude said, you're banned. Kicked out of the military, gone just like that. And within hours, OpenAI swooped in and took the deal.

That's not normal. That is a war. But that's not even the craziest story this week. A brand new AI tool just launched called Plexi computer.

Here's what that means in plain English. Instead of using one AI to help you, this thing uses 19 different AIs at the same time, all working together on the same task. It's like having an entire team of AI workers instead of just one. And it runs for hours, days, even months without stopping.

Then there's OpenClaw. If you've never heard of it, it's the fastest growing AI project in the world right now. And this week, some happened with it that you need to know about. There's also breaking news from Google, from Anthropic, from Notion.

And there's a story about China dropping an AI tool so powerful that Hollywood studio sent legal letters in a panic. I'm going to walk you through every single one of these stories today. No tech decree needed. No jargon, no fluff.

Just what happened, why it matters and what it means for you. Because this stuff is moving fast. And if you're not keeping up, you're falling behind. Let's get into it.

So let's talk about the day the US government banned Claude. I want to start here because this is a story that's going to affect everything. Every company, every developer, every person who works with AI in any professional setting needs to understand what just happened. Anthropic is a company in San Francisco.

It was founded by a man named Dario Amodi and his team, many of whom used to work at OpenAI. Anthropic makes an AI called Claude. Claude is the AI tool I often use. And it's the AI that a lot of the most complicated and sophisticated companies in the world use.

And until this week, Claude was the only AI system cleared to run inside the US military's classified computer networks. That means the secret ones, the ones handling intelligence operations, the ones running national security decisions, the ones that were reportedly used in the mission that captured Venezuelan President Nicolas Maduro. In July of 2025, the Pentagon signed a contract worth up to $200 million to use Claude. That's a $200 million deal.

For context, Anthropic is valued at $380 billion. So this contract was not existential for the company. But what happened next was potentially much bigger than the contract. Here is what went down.

The Pentagon came to Anthropic and said, we need you to remove your rules. Specifically, they wanted Anthropic to allow Claude to be used for two things. Number one, mass surveillance of American citizens. Things like, for example, tracking location data, web browsing history, personal financial information, and number two, fully autonomous weapon systems.

That means weapons to fire on their own with no human making the final decision. Anthropic said no. Now, here's the thing. Anthropic had rules against those two things baked into Claude since June of 2024.

The rules were not new. The Pentagon knew about them. They agreed to the contract with those rules in place. But something changed.

Defense Secretary Pete Hegseth met with Dario Amodi this week and delivered what he called the best and final offer. Remove all restrictions on how the military uses Claude. Allow Claude to be used for all lawful purposes. No exceptions.

Dario Amodi wrote a letter in response. He said, quote, we cannot in good conscience accede to their request. That letter ended the relationship. Hegseth went on social media and announced Anthropic would be designated a supply chain risk.

That is a label normally used for companies that are connected to foreign adversaries. Companies like Huawei, the Chinese tech firm that the U.S. government has been fighting for years. They gave that same label to Anthropic, a U.S.

company valued at $380 billion, founded by Americans built on American soil. Then President Trump posted on True Social. He wrote, I'm directing every federal agency in the United States government to immediately cease all use of Anthropic's technology. We don't need it.

We don't want it. And we will not do business with them again. And just like that, Claude was banned. Federal agencies have six months to remove Claude from their systems.

Any company that does business with the U.S. military now has to certify that they are not using Claude in that work. Palantir, one of the biggest defense technology companies in the world, uses Claude to power some of its most sensitive military work. They now have to find a replacement.

And here's the kicker. Defense officials in private conversations reported by Axios called it a, quote, huge pain in the ass to disentangle from their existing systems. Claude is that good. It is that embedded.

They know it and they are doing it anyway. Now, let us be clear about something, because it's always good to look at to look at everything. So we've got the Pentagon's argument that is straightforward. You know, when the government buys it all, they should have full control over how they use it, and they're not saying they want to spy on Americans or launch robot wars or something like that.

They're saying that a private company should not get to put limits on what the government can and cannot do. We've legally purchased technology. That's not a crazy argument, right? At the same time, Anthropix rules against mass surveillance and autonomous weapons were publicly stated.

They were in the terms of service, the Pentagon read them and agreed to them, but they just changed their mind and demanded something different. So it's an interesting one, and this is what makes it so unusual. And here's a question everyone is asking. What happens to Anthropix?

The short answer is it hurts, but probably not as much the government might think. The 200 million dollar contract itself is not existential for a company valued at 380 billion dollars. The bigger danger is the supply chain risk designation, because now any company that has Pentagon contracts has to prove they are not using fraud. And a lot of big companies use fraud.

If those companies have to choose between their Pentagon business and their fraud subscription, many of them will choose the Pentagon. And that's the part that could really bite. On the other hand, something interesting happened. The moment the ban was announced, hundreds of engineers at Google, Microsoft and Amazon signed internal letters demanding that their companies take the same position as Anthropix.

They do not want their AIs used for mass surveillance either. So Anthropix took a 200 million dollar hit and possibly gained the moral high ground in one of the most important debates in the history of technology. Is this the best thing that ever happened to the Anthropix brand? Some people might say yes.

When a company stands on principle and loses money for it, people notice. We'll all be watching this closely. Within hours of the Anthropix ban being announced, Sam Altman, the CEO of OpenAI, posted on social media he had reached an agreement with the Pentagon. Chachi Petit was coming to the military's classified systems to replace Claude.

Now, here's the part that made a lot of people raise an eyebrow. Sam Altman had told his employees in an internal memo just one day earlier that OpenAI would push for the same restrictions. Anthropix has the same restrictions. No mass surveillance, no autonomous weapons, same red lines.

And then within 24 hours, a deal was done. What were the actual terms? Nobody is saying that clearly. The public statements are a little bit vague.

This is the part I want to be honest about because I don't want to pretend I know something I don't, right? What we know is that OpenAI got the deal. What we do not know fully is exactly what was and what was not agreed to. What we do know is that ex-AI Elon Musk's AI company has already signed a deal to be used in classified settings and that Google's Gemini and OpenAI's Chachi Petit are both in conversations about the classified space.

The world's most sensitive military AI work is now up for grabs, and whoever wins that business will have a relationship with the most powerful government around, right? So this is not a small thing. This is worth paying attention to. We're going to shift gears now and talk about Perplexity Computer because whilst the Claude ban was dominating the news, something else dropped that I think is actually more important for most people's day to day lives.

Perplexity launched something called Computer. If you've not heard of Perplexity, here's a one sentence version. They are an AI search company. Instead of using Google and getting a list of links, use Perplexity and get an actual answer with sources.

They grew their users by 3.7x in 2025. They grew their revenue by 4.7x. And now they're trying to become something much bigger than a search engine. Computer is a product that costs $200 a month for the highest tier.

And what it does is pretty well. You give it a job, one job, something like, for example, research my three biggest competitors and build me a website with the findings or manage a full marketing campaign for this product launch. And then Computer goes to work by itself without you doing anything. Here's what makes a difference from everyone else out there.

So Computer does not just one AI to do the job. It uses 19 different AI models. Think about that, 19. For example, Claude opens 4.6 as its main thinking engine.

Gemini for deep research, NanoBanana for creating images, Veo 3.1 for video, Grok for quick searches and ChatGPT for long documents. It assigns each part of your task to the AI that is the best at that specific thing. The CEO of Ploplexy, a man named Aravind Srinivas, said something that stuck with me. He quoted Steve Jobs.

Musicians play their instruments. I play the orchestra. That is what Computer is doing. It is conducting an orchestra of 19 different AI models.

And you just tell it what song to play. Now, here's the honest part, because I do not just like to give you marketing language about the full picture. There's a real risk here. Ploplexy does not own a single one of those 19 models.

Every single AI it uses belongs to a company that either competes with Ploplexy directly or is building their own version of the same thing. If Anthropic raises its prices, Computer gets more expensive to run. If Google decides to prioritize its own age and products, they can make Gemini harder to access. Ploplexy is building a product on top of other companies' tools.

And those companies know it. That said, the vision is real and it's big. The era of using one AI for everything is ending. The era of AI systems that route your tasks to the best tool for each.

job is just beginning. But Blexi is betting that the company that builds the best orchestration layer, the conductor of the orchestra, wins even if they do not own the instruments. That's a real bet and it might be right. I need to tell you this story next because it's one of the most interesting things that happened this year and most people have not heard the full version.

Peter Steinberger is an Austrian software developer in his 40s. He built a company called PDF Kit, a PDF toolkit that runs inside apps like Apple, Dropbox, SAP, IBM and the European Patent Office. Nearly a billion people have used his code without knowing it. He sold his shares when investors put in 160 million dollars in 2021.

He retired and for five years he barely touched a computer. And then one day in November 2025 he decided to tinker. He built something called ClaudeBot. It was a personal AI agent that ran on his computer, not in the cloud, on his machine with access to actual files, his actual apps, his email, calendar and smart home devices.

The agent could do real things, not just answer questions. It could book flights, clear his inbox, control the temperature of his bed even. You talk to it through WhatsApp or Telegram like texting a very capable assistant who also happened to have the keys to your house. And he put it on GitHub, the website where developers share code.

And then something happened that surprised even him. People went crazy for it. The project got 100,000 GitHub stars in 10 days. For context, that is more stars than projects like Next.js that took years to build and are used by millions of developers.

Within weeks the name changed. First to MoldBot, after Anthropic sent a trademark letter because the word ClaudeBot was too close to Claude. Then to OpenCore and the stars kept climbing. 160,000 GitHub stars, then 108,000, then 200,000 and 241,000 right now.

The fastest growing open source project in the history of GitHub. Both Meta and OpenAI came knocking with offers. Mark Zuckerberg reached out personally via WhatsApp. Sam Altman made his pitch.

Peter's condition was simple. The project has to stay open source. It cannot go behind a paywall. It cannot be locked inside any company's products.

He chose OpenAI. Sam Altman announced on February the 15th that Peter was joining to quote, drive the next generation of personal agents. OpenCore moves to an independent foundation that OpenAI will financially support. The code stays free for everyone and here's the part that I think most matters.

So Peter built this project primarily using Claude. He named the original version after Claude. He was one of the biggest drivers of API revenue going into Anthropic's bank account and Anthropic sent lawyers. The trademark letter that forced the first name change may have been the thing that pushed Peter towards OpenAI when the bidding war started.

One of the smartest observers in AI called it a masterclass in knowing the difference between protecting your brand and alienating your ecosystem. OpenAI meanwhile hired the creator of the fastest growing AI project in history, kept it open source and attached themselves to a community of hundreds of thousands of developers who are already obsessed with this thing. For essentially nothing compared to what Meta reportedly offered. I keep thinking about what Peter said about where all this is going.

He thinks 80% of current apps are going to disappear. Why would you need a separate weather app or a separate flight booking app or a separate reminder app when one agent that knows you, knows your calendar, knows your preferences can just handle all of it through a text message? That is not science fiction. That is the direction and OpenClaude now with 241,000 GitHub stars and OpenAI's resources behind it is one of the clearest signs of where we are heading.

Before we go any further I want to take a second to talk about the AI profit boardroom. If you're listening to this podcast and you're trying to figure out how to actually use AI in your business, in your career or to build something the boardroom is where I break it all down in the community of people doing the same thing. Every week we go deeper than the headlines, we look at what these tools actually do in practice, we share what's working, what is not and what is coming next and if you want to join us go to the link in the description. The community is growing fast and I think you will find your people there too.

It's available link in the comments description or go to the AIprofitboardroom.com. Okay back to the news. Cdance 2.0 the AI video that freaked out Hollywood. ByteDance makes TikTok, most people know that.

What fewer people know is that ByteDance also has one of the most aggressive AI research programs in the world. They have been quietly hiring top engineers away from Google DeepMind, Alibaba and every major AI lab. They have spent billions building out their infrastructure and on February the 8th of this year they dropped something called Cdance 2.0. Here's what it does.

You type a prompt or upload an image or a video clip or an audio file and Cdance creates a video up to 15 seconds long with synchronized audio with realistic physics meaning things fall all the way they're supposed to form or fabric that moves the way fabric is supposed to move. Explosions have actual weight with cinematic camera work meaning dolly zooms, tracking shots, slow motion, all of that. These are all generated automatically in 2k resolution and the results were so good that things started going viral immediately. Someone posted a 15 second clip of what appeared to be Tom Cruise fighting Brad Pitt generated from a two-line text prompt.

Rhett Reese, one of the writers of Deadpool and Zombieland saw it and wrote publicly, I hate to say it, it's likely over for all of us. He went further he said that in not much time one person sitting at a computer will be able to create a movie indistinguishable from what Hollywood releases right now. That is a professional Hollywood screenwriter saying this. Disney sent a cease and desist letter to ByteDance Paramount.

Someone as well and the Motion Picture Association put out a statement calling what ByteDance did quote unauthorized use of US copyrighted works on a massive scale. SAG after the actors union condemned it and ByteDance responded by disabling the feature that let people copy real celebrity likenesses. They said they would strengthen safeguards but here's the thing the damage or depending on how you look at the demonstration was already done. The video of Cruise fighting Pitt had already gone around the world.

People had already seen what this tool could do and this is a pattern we're going to see over and over again. A Chinese AI lab released a tool with fewer restrictions than US companies are allowed to have. It goes viral, the legal letters come, some features get disabled but the cat is out of the bag. This is the second time in 2025 and 2026 that a Chinese AI has caused this kind of shock moment.

The first was DeepSeek which proved you could build a world-class AI model for a fraction of what US companies were spending. Now SeedDance is proving that one person with a $60 subscription and a good idea can produce video that would have taken a major studio about six months and millions of dollars to create. And I'm not here to tell you whether that's good or bad but I will tell you it's real and it's accelerating. Now let's talk about Nano Banana 2.

Google's image AI gets faster and smarter. So Google have a funny name for their image generator they call it Nano Banana. You might laugh at that name but the results are not funny at all. The original Nano Banana was launched in August of 2025 and went completely viral especially in countries like India where millions of people were generating images in the Gemini app every single day.

Nano Banana Pro came out in November and now this week Google dropped Nano Banana 2. Here's what changed. Nano Banana, technically called Gemini 3.1 flash image, combines the higher quality output of the pro version with the speed of the faster flash model. That means you get better images faster.

It can create images up to 4k resolution. It can keep up to 14 objects consistent across multiple images which is huge for creators making comics storyboards or any kind of visual storytelling. But the thing I think is the most new interesting feature is something they're calling real-time web grounding. The model can pull actual current information from web search whilst it's generating an image.

They built a demo called WindowSee. You tell it a location anywhere in the world it pulls the actual current weather there and it generates what you would actually see looking out a window from that specific place at that specific time of day of that specific weather. That is not a static image database. That's a model that understands the world in real time and renders it visually.

Nano Banana 2 is now the default image model inside the Gemini app. If you use Gmail, Google Docs, Google search or any Google product with AI features you're probably now using Nano Banana 2. And according to third-party benchmarks Nano Banana 2 beats OpenAI's best image model on the major leaderboards at roughly half the price per image to generate. This is Google showing up.

This is what I keep telling people. Google was slow to react to Jack Cheapity in 2023. They looked flat-footed and people wrote them off. Don't write off Google.

They have the world's best search infrastructure. They have DeepMind. They have YouTube. They have the largest distribution network for AI features in the world and Nano Banana 2 is not a trend.

It's Google playing for keeps. Google Flow. One place for all your creative AI. So once we're talking about Google, let me tell you about three more things they dropped this week because it was a busy week for them.

The first is Google Flow. Flow launched last year as Google's AI filmmaking model powered by their video AI called Veo. People have already used it to generate over 1.5 billion images and videos. That's 1.5 billion with a capital B.

In a few months, this week, Google made a major move with Flow. They took two of their other experimental image models, one called Wisk and one called ImageFX, and they folded them both directly into Flow. Now, Flow is not just a video tool. It's your one-stop shop for all creative AI work.

Make an image, edit it with a lasso tool using plain English commands, turn it into a video, add audio, manage all your projects in one place. The new interface has an asset grid where you can sort, filter, and tag all your work. And you can use the add symbol to quickly reference specific files in your prompts. With the new version of Veo, they are 3.1, you get richer, better camera control, better audio and more cinematic realism across the whole experience.

Starting in March, if you use whisk or image effects, you can move all your old projects directly into flow, nothing gets lost. This is what good product design looks like. Instead of making you use 5 different tools or do slightly different things, Google is saying come here, we'll put everything in one place and it will work together. For marketers, content creators, educators, anyone who makes visual content, this is worth paying attention to.

The second Google thing I want to cover is Opal. Opal is a tool from Google labs, which is where Google tests experimental ideas before they go to the main products. Opal lets you build what they call mini apps. These are small AI powered tools that do a specific thing automatically.

You do not need to know how to code to use it. You describe what you want in plain English, AI builds it for you. And this week, Google added what they call an agent step to Opal. Before this update, you would build a workflow in Opal and it would follow a fixed path.

Do step one, then step two, then step three. With the new agent step, you set a goal instead of a path. You say, I want this agent to find the best answer to this problem. And the agent figures out the path itself.

It picks his own tools, it routes his own decisions, even ask you clarifying questions when it's not sure what you want. They also added memory, meaning the AI agent can remember things from previous sessions. So if you have a customer service agent and a customer calls for the third time, the agent already knows the context from the first calls. Opal also now connects directly into Google Workspace.

That means Gmail, Google Docs, Google Sheets, Google Calendar. Your agent can read your email, update your spreadsheet, schedule a meeting, all without you switching apps. VentureBeat covered this update and called it the new blueprint for enterprise AI agents in 2026. That is a bold claim, but I think they're right.

The shift happening now is from AI that answers questions to AI that actually does things. Opal and the agent step update specifically is one of the clearest examples of what that shift looks like in a real product that real people can use today. Alibaba, Quen 3.5, China's open source AI that keeps getting better. China's biggest tech company is Alibaba and the AI team called the Quen team has been on a serious run.

This week, timed perfectly for Chinese New Year, they dropped Quen 3.5. Now, I want to explain this in plain terms because there's a lot of technical jargon around these model releases and most of it is just noise. Here's what actually happened. Quen 3.5 is open source.

That means you can download it for free, run it on your computer, build on top of it, fine tune it for your business and all without sending a single piece of data to Alibaba's servers. You can use Alarma or whatever you want to run this. The model has 397 billion parameters which is a technical way of just saying it's very large and very capable. But here's a trick.

It only activates 17 billion of those parameters at any given time when it's actually running. Think of it like a restaurant with a kitchen staff of 397 people but only 17 people are cooking at any one time and they're the exact right 17 people for whatever dish is being ordered. That architecture makes it extremely fast to run and extremely cheap to operate. Benchmark tests, and I will note these are partially self-reported, show Quen 3.5 performing on par with the best models from OpenAI, Anthropic and Google.

On some specific tasks it actually outperforms them. On many it's in the same range. The Quen 3.5 Medium series, released just a few days later, includes a model that on third-party benchmark tests actually beats OpenAI's GPT-5 Mini and Anthropic's QuadSonic 4.5. And it's free.

You can run it on a single high-end GPU. The context window, which is how much information the model can hold in its head at once, goes up to 1 million tokens on the hosted version. That's one of the largest in the industry. The model also supports 201 languages, up to up from 82 in the previous version.

And this is what I keep saying when people ask me about the US versus China AI race. The gap is not as large as a lot of American commentators want to believe. It's close and fast. Quen 3.5 is not just a research achievement, it's a production-ready, enterprise-grade, open-source AI that any company in the world can use right now for free.

For smaller companies that cannot afford the API cost of quality GPT, this is a life-changing unlock. For enterprises worried about data privacy and sovereignty, a model you can run on your own hardware is enormously attractive. And for the global AI race, this is a reminder that the competition is not slowing down, it's speeding up. Now let's talk about something more practical.

Something you can actually start using this week. Notion released custom agents on February 24th as part of Notion version 3.3. If you've never used Notion, the one sentence description is a workspace tool where teams write documents, manage projects, and organize information all in one place. Millions of companies use it.

Tens of millions of people use it every day. And now these people can create AI agents that work inside the Notion workspace automatically without being told to do anything. 24 hours a day, 7 days a week. Here's how it works in plain terms.

You describe a job. So for example, I want an agent that monitors our Slack channel for customer questions and answers them automatically using our Notion knowledge base. Or I want an agent that reads incoming support tickets, routes them to the right member, and sends an acknowledgement to the customer. Or I want an agent that every Monday morning compiles the status of all active projects and sends a summary to the whole team.

You set the trigger, maybe it's a message in Slack, a new entry in a database, or just a schedule like every day at 9am, and the agent runs by itself without you. Early testers built over 21,000 custom agents during the beta period. Notion itself runs 2,800 agents internally. More agents than the company has employees.

Let that land for a second. Notion has more AI agents doing work than they have human workers. James Lawley, the IT operations manager at a company called Remote, saved his team 20 hours a week using custom agents to triage and route support tickets. The agents triage with more than 95% accuracy.

They resolve over 25% of tickets completely on their own. A team at RAMP, the financial software company, uses custom agents to answer dozens of nuanced product questions every day. Questions that used to require a human to look up the answer and respond manually. Now the agents do it.

The human team just audits the responses and feeds improvements back in. This is not a vague future prediction. This is happening right now in real companies with real results. Custom agents are free to try through May 3rd, 2026.

After that, they require Notion credits, which are an add-on for business and enterprise plans. If you're on a team that uses Notion, I would go and set one up this week. Just see what it feels like to hand off a job to something that does not need to sleep. And the last thing I want to cover today is something that might sound small compared to everything else we've talked about.

But if you use Notebook LM, you have been waiting for this for a long time. Google's Notebook LM is one of my favorite AI tools. And the way that it works is you upload documents, notes, videos, web links, whatever you have. And Notebook LM reads all of it and becomes an expert on your specific material.

Then you can ask it questions, generate summaries, and even have it create a podcast style audio conversation about your topic. It also generates slide decks. Beautiful. Professionally designed slide decks based on your source material.

The problem up until now was that once Notebook LM made those slides, you could not edit them. Every slide was basically a flat image. If you wanted to change one word, you have to generate the entire deck from scratch. That just changed.

Google rolled out what they call the revise feature earlier this month. You open your deck in the studio panel. You click the pencil icon. Each slide gets its own text box where you can type instructions directly to the AI.

Change the color of the slide. Replace the image with something more professional. Add a statistic here. Simplify the text.

And you can make changes to multiple slides at once. They get grouped together as pending changes. And then you can click generate new deck and it applies everything. They also added PowerPoint export meaning you can download your Notebook LM slides directly as a PowerPoint file.

Google Slides export is coming next. This update rolled out to all Notebook LM users, not just paid subscribers. So free accounts can get it too. There are still some limits.

You cannot add or remove slides yet. You can only modify existing ones. And the PowerPoint files are still rendered as images inside PowerPoint rather than fully editable text boxes. So you cannot click in and just retype something directly in PowerPoint.

But this is a significant step forward. Notebook LM is already one of the most useful AI tools for anyone who has to turn a lot of information into something other people can understand. Students, researchers, marketers, consultants, educators. The ability to customize slides it creates rather than take whatever it gives you makes it dramatically more useful for real work.

So let me step back and tell you what I think all of this means. Because if you look at everything we just covered as separate stories, you might miss the larger thing that's actually happening. The Claude Ban tells you that the fight over who controls AI is no longer just between companies. It's between companies and governments.

And the outcome of that fight is going to determine what AI can and cannot be used for in the most consequential settings including war. Anthropics stood on principle and lost 200 million dollars but they may have won something more valuable. The trust of every enterprise customer who's watching and thinking this is a company that will not just hand over the data to whoever asks. The OpenClaude story tells you that the next billion dollar AI product might be built by one person in an afternoon.

Not because the person is a genius but because the tools are now good enough that a retired engineer can build the fastest growing projects in GitHub history whilst also having thoughts about whether his hands are too precious for typing. 241,000 stars person, no venture funding, no team, that is where we are. Plexi Computer and Notion Customer Agents are telling you the same thing from two different angles. The future of software is not more apps, it's agents that coordinate other agents.

Systems that can be given a goal, not a list of steps, and figure out how to get there on their own. We're still in the early, rough, sometimes chaotic version of that future. Open core instances were getting hacked, but Plexi does not own the 19 models it orchestrates. Notion's agents can occasionally do the wrong thing, all of that's real, but the direction is clear.

Cdance 2.0 and Hollywood Panic tells you that the content industry is about to experience what the music industry experienced with Napster, except faster and with higher stakes. One person, one promise, one tool that costs the same as a streaming subscription, with Hollywood quality output. The question is not whether this changes the entertainment industry, it already is. The question is really, what is the new structure going to look like once the dust settles?

And the wave of Chinese AI, Quen 3.5, Cdance, DeepSea before them, tells you that the AI race is truly global. Alibaba is releasing world class models for free, ByteDance is releasing video generators that make Hollywood studios nervous. The US companies are not running away from competition, they're in a real race. And the outcome of that race has implications for business, for jobs, for national security, and for who gets to decide how the most powerful technology in history is used.

That's not small, that's the defining story of this decade. So what should you actually do with this? I know a lot of you are listening to this and thinking, okay, this is fascinating, but what do I actually do with it? Fair question.

If you're a business owner or manager, go look at Notion custom agents this week. Seriously, not to replace your team, but to find the most repetitive tasks your team does and see if an agent can handle them. The company remote saves 20 hours per week on one workflow. What with 20 hours per week free up for your team.

If you create content, presentations, reports, or educational materials, go try the new Notebook LM slide, editing, put your source documents in, generate a deck, then use the revise feature to make it match your brand and your message. This is a workflow that used to take hours, it now takes minutes. If you work in tech or you're building something, pay attention to OpenCore and the agent ecosystem growing around it. 241,000 stars means 241,000 developers who think OpenCore is the future.

The tools, the patterns, the architectures coming out of that community are going to shape what the next generation of software looks like. And if you work in or near the entertainment industry, whether that's film, TV, advertising, gaming, visual design, I would encourage you to try Cdance 2.0 yourself. Not to panic, but to understand what you're actually dealing with. The Deadpool writer said it's likely over for us.

I'm not going to go that far, but I would say the creators who understand these tools and use them as leverage will have a very different future than the creators who refuse to look at them at all. The last thing I will say is this, every week when I sit down to cover this stuff, I think about the people who are caught in the middle. The person who has been a paralegal for 15 years and is watching AI handle legal research faster than they can. The animator who spent a decade learning their craft and is seeing AI generate in seconds while you stay weeks.

The customer service rep who is watching their company build AI agents to do exactly what they do. These are real people and this is a real transition. I do not think the answer is fear, but I also do not think the answer is just optimism. The answer is engagement.

Learn the tools, understand what they can and can't do, find the places where your judgement, your relationships, your creativity, your humanity are still irreplaceable, and build your future there. Because the tools are not going away and the pace is not slowing down. If anything, this week proved that it's speeding up. I will see you next week and if you want to go deeper on any of this stuff, come find me and my community inside the AI Profit Boarding.

Link in the comments description or go to theaiprofitboarding.com