AI Agents Are Building Entire Companies While You Sleep

The most surprising thing has been the lack of public recognition of how close we are to the end of the exponential. This is absolutely wild. At this point, everyone in the AI industry is ringing the alarms. If you haven’t been following along, there’s a lot of people saying the same thing: you need to wake up. Dario Amade, the founder of Anthropic, is saying that we’re at the end of that exponential. We’re approaching, we’re near the end.

The Wake-Up Call Everyone Ignores

Recently, one of the AI safety researchers at Anthropic left writing a fairly alarming message to his co-workers, saying the world is in peril. Jimmy Ba, one of the original founding researchers at XAI, said it’s his last day, and he’s saying recursive self-improvement loops are likely to go live in the next 12 months. 2026 is going to be insane and likely the most consequential year for the future of our species.

Remember back in early 2020? There was this subreddit calledflu named after some city in China. A lot of people in that subreddit were kind of predicting what was about to happen. This was when the rest of the world had no idea. And then slowly you saw people waking up almost one by one. If you just woke up to that fact when you noticed that the toilet paper was out of the stores, you were late to the party.

Right now, we’re actually entering the new era. The world out of toilet paper era, or whatever the analogous thing is for AI, and I think that thing is actually AI agents.

I Built an Entire Company While Sleeping

I am about to unveil the thing that I’ve been building with it. It’s built entirely with AI agents with OpenClaw specifically. Watching them build out the entire thing was kind of nuts because I gave them the big idea. They came up with the entire buildout, what technology to use. I set them up on GitHub. So they’re pushing various updates to GitHub. Then they’re deploying it on Vercel.

I know very little about actually how to do all the technical things. I just sit there on my phone texting it, and this thing builds out entire software suites that are needed to bring that vision to reality.

There was a very good post on X by Matt Schumer. It’s sitting at 80 million views now, absolutely blowing up. It’s called “Something Big Is Happening.”

The first thing that he said is, “I know this is real because it happened to me first.” Here’s the thing that nobody outside of tech quite understands yet: the reason so many people in the industry are sounding the alarm right now is because this already happened to us. We’re not making predictions. We’re telling you what already occurred in our own jobs and warning you that you are next.

The Test Is No Longer Can It Write Code

I’ve been testing these AI models since 2023. In the beginning, they were simple word puzzles, riddles. Write this thing out for me, and often they would fail. Then over time, we transitioned to coding. Can’t this thing create a little snake game? Eventually, it could. But each new model release kind of got better and better.

But in this last week with the release of GPT-5.3 Codeex and Opus 4.6, this idea of prompting it and then seeing what it comes up with as a test of its abilities, it doesn’t even make sense anymore.

Because really the real test is when you hook it up to something like OpenClaw, for example, and then give it a big idea and then tell it to just run through the night building that idea out and deploying it. Then you wake up in the morning and you look at the output, and the output is what used to be a business, a company.

So the test is no longer “can it provide a chunk of code” or “can it troubleshoot this one isolated detail.” The test is now: can it build a business overnight, or at least the tech side of it.

28 Commits in Under 4 Hours

On February 8th, 2026, technically day zero because I had this idea late on that Sunday and before going to sleep, I told it to build it out. So it called it “The Build Begins. Night Sprint.”

Between 9:50 p.m. and 1:30 a.m., the AI agent builds the entire platform from scratch. 28 commits in under 4 hours. First deploy to Vercel goes live.

Do you want to know what I was doing between 9:50 p.m. and 1:30 a.m.? I was asleep. I told it what to do. I kind of said, “Okay, like do you have a rough plan outlines?” Like, “Yeah, we’re going to be kind of doing this.” I’m like, “Okay, make it.” So, number one, and then I went to sleep. In the morning, the thing was live and functional and looking good, actually working.

The Centaur Age

In the interview with Ilon Musk, he was saying how this might be one of the things that is missing from these models because us as humans, we tend to chase some state. We tend to imagine that, oh, if I get that promotion or if I build that business, we tend to envision that state and think that’s going to make us happy, and then we sort of chase that state, and that’s kind of what allows us to agentically move forward towards that state even if we don’t know the exact path.

So we don’t have something like that for these AI agents. They’re kind of like pretty zen and chilled. They don’t have this drive. You have to push them. But all that capability to take those next steps and pursue medium-length tasks, medium-horizon tasks, it’s all there. We’ve just turned that corner.

Dario Amade refers to that as kind of like the centaur moment. The whole idea of a centaur is half man, half horse. And this is as opposed to a cyborg. So in the context of AI, a centaur is like the AI does something and then you step in and you do something and you hand it back and forth in order to complete the task.

The End Game

So, yeah, I think we’re entering the singularity. I think we’ve got a few years ahead of us, I’m guessing, during which I think we’re going to see some crazy businesses being built by autonomous agents, these agentic enterprises, autonomous enterprises. And then after that, something I don’t know, we’ll see. But yeah, feels like we’re approaching the end game, whatever that is.

If you’ve ever wanted to build a business, if you’ve ever had an idea that you think is useful, now is the time. It might be the best time in the history of the world to build it. And maybe these next few years, this might be the last time that you could build something like it because again, if you project into the future, at some point you can imagine being able to say one sentence to your AI agent and then it creates the entire Adobe suite if you wanted it to.

The software economy is about to undergo the most dramatic transformation in its history. Just as the internet fragmented entertainment into millions of personalized streams, AI will fragment software into billions of personalized agents. Photoshop will be broken into a million pieces, each for its own use case.

The question isn’t whether this future is coming – it’s whether you’ll be ready when it does.