On August 23rd, a lot of people are going to lose eight months of work.

Not because they did anything wrong. Not because anyone got hacked. Not because they missed a payment or violated a policy.

Because two governments got in a fight over who owns the company they were building on.

The company is Manus. If you were paying attention to AI last year, you know it. It was the thing that made people sit up. You gave it a job and it went and did the whole job, browsing, writing, building, finishing. For a lot of business owners it was the first time AI stopped feeling like a chatbot and started feeling like an employee.

It was built in China by a startup called Butterfly Effect. In June of 2025 they moved their headquarters to Singapore. Chip access, investor pressure, distance from Beijing. Then in December, Meta announced it was buying them for around $2 billion.

China said no.

Not right away. First they opened an investigation. In March, both co-founders were barred from leaving mainland China while it ran. Then in April, the state planner ordered the whole thing unwound. Not blocked. Unwound. The deal had already closed. Investors had already been paid. Employees had already walked into Meta.

Reverse it anyway.

This month that order finally reached the people who were just trying to run their businesses. Anything created in Manus on or after December 29th, 2025 gets deleted. Accounts go dark for two days. There's a backup window, and if you're outside of it, that's the end of the conversation.

Manus isn't the villain here. They got picked up and moved by something much bigger than them, and everyone standing on top of them got moved too.

I left in April and told everyone I could to do the same.

I want to be honest about why, because it wasn't foresight. I didn't predict any of this. I'm not that good.

It was simpler than that. I noticed that everything I made in there stayed in there.

Months of context. The way it had learned to think about my business. The instructions, the corrections, the accumulated understanding that made it useful in month four in a way it wasn't in week one. All of that was real value, and all of it lived on a server I had no claim to.

I couldn't answer a basic question: if this disappeared tomorrow, what would I still have?

Nothing. I'd have nothing.

That's not a prediction. That's just reading your own position honestly.

Here's the part I think people get wrong. They hear this and assume the lesson is about China, or about picking safer vendors, or about reading the news more carefully.

It isn't.

The lesson is that there's a difference between using someone's models and living in their house.

I still use their models. Claude, ChatGPT, Gemini, all of them. I have no interest in training my own and neither should you. That's not the ownership that matters.

What matters is where the work lands.

Mine has three pieces, and none of them are exotic.

An interface. I use Conductor. It's an app that runs on my machine and connects to whichever models I want working. It's the screen I look at. It's also the piece I could throw away tomorrow and replace by Friday, which is the point.

A repo. This is the actual brain. Every file, every decision, every instruction for how the system should think about my business, stored in a Git repository I control. I could lose my laptop today and pick up exactly where I left off. Claude, Codex, Gemini and Grok all read from it and write back to it. A team can share one. Each of my clients has their own, and I hold the tools and agents in mine.

Connections. Gmail, YouTube, Meta, Cloudflare, Drive, Telegram, about forty of them. Same capability people loved about Manus, connect the AI to your stack and let it actually do things. The difference is the credentials live on my machine instead of in a company's database.

That's it. It isn't clever. It just means the models are interchangeable and the work isn't.

I can switch from Claude to something else this afternoon and lose nothing. If a company I depend on gets acquired, repriced, blocked, or shut down by a government I've never thought about, I lose a tool. I don't lose my business.

I've been making this argument for months, and the honest problem with it was always that it sounded theoretical. Nice idea, Mike. Sounds like a lot of work for a risk that isn't real.

It's real. It has a date on it now.

The platform doesn't have to die for you to lose everything. It doesn't have to get hacked or go bankrupt or turn evil. It just has to change hands, and you don't get a vote in that. You're not a shareholder. You're not a party to the transaction. You're a line item in someone else's unwind.

And you find out on a Tuesday.

If you're building inside a platform right now, you don't need to panic and you don't need to rip it out this week. You need to answer one question honestly.

If this was gone tomorrow, what would I still have?

Whatever the answer is, that's what you actually own. Everything else was borrowed.

If you're not sure where your own business is exposed, I built something for that at bledsoe.life/start. It asks you a handful of questions and finds the constraint that's actually costing you. For some people the next step is the community. For bigger businesses it's a conversation with me. Either way you'll leave knowing which one you are.

Now read your own business the way you just read this one. Run the Check >_

Or get the writing, once a week.

New essays on ownership and the AI shift.

← All writing