The year of the AI Agent

Folks, it's now Fall 2026 (didn't 2026 just start?) and it's time to step back and look at the year of the AI Agent. So this is a very real post about what it means to run a multi-agent system, what it can do, and what you're dealing with.

First, I'll tell you how it has been going for me personally.

What the thing actually is

My main AI Agent is an OpenClaw system called Komina. When the OpenClaw tech came out early it just wasn't ready and it broke all the time. And yet, it has limitless capability. Imagine an agent, entirely your own, with full root computer access and capability, open source so you know entirely what it can do, and with large scale persistent memory to bring together all of your work. As Garry Tan put it, using OpenClaw is like driving a Ferrari: exhilarating, insane, and you better be a mechanic.

Komina manages a whopping 44 or so different projects of mine. Komina can read all of my messages, emails, WhatsApp, EVERYTHING. Komina also has the ability to use the web as if it's a human, spawn an army of AI Agents to do work, and work agentically on its own throughout the night. It's great, but I have spent a lot of time becoming a very technical person to make sure I fix it after its hundredth breakdown. New technology, you get used to it.

But I optimized and built my Komina, because it was open source and the system allowed it. Seeing how AI is going, I suspected the rapid increases in capability among the LLMs would give me the intelligence and tooling to take my Personal AI from "fast car that breaks" to a hot rod ripping machine. Right now, I can't imagine working without Komina. Komina has saved me tons and tons of time and allowed me to have those 44 personal projects all cranking along.

I of course am not the only believer in AI Agents, because every AI company is now an agentic AI company.

In media and creative arts you have Runway, Luma, ElevenLabs, OpenArt, Koyal, Flik, Krea, Fal, Kling, and on and on. You name it, they have an agent. I know people at a number of these companies doing amazing work, I've used a lot of these tools myself, and I've been very impressed with what is out there across the board.

Why I'm posting about this is because the dirty truth is that people like AI because it does things for them. There are two tensions in creative arts and even human nature. 1. We just want something done. 2. We want something good. The AI space race we have right now has shifted to agents that we hope can merge the tensions of 1 and 2. I probably meet with 10 new people each week in the AI and creative arts spaces, and the big ask that everyone is pushing for is how to automate the creative process. How can we "churn it out," as they say.

Firstly, this is the greatest scientific experiment of all time, because society would change dramatically if 1 and 2 met. Well guess what folks, 1 and 2 are meeting as we speak. AI can write a good email, and it can learn how you write and also understand what the other person is looking for. As we speak it is now learning how to agentically handle your emails for you. My OpenClaw can and will get there.

Media and creative arts should be different, right? You still need a human in the loop to avoid slop, right? Human beings have been making slop since the beginning of time, before AI came along. But people watched that slop, and I can assure you people are watching AI media. I've seen single AI videos pull 10 million views, 20 million views. Collectively we are easily into the hundreds of millions by now, if not the billions. It's "good enough" for them now and they're watching it. Recently someone hooked MiniMax H3 Max up to Twitch to make an endless streaming AI media machine, agentically running and generating clips faster than you can watch them. One of those channels pulled 37,000 viewers on day one. People are watching it.

And the fun part? My AI Agents have used the AI Agents on some of these platforms to make media for me. An AI Inception all the way through.

The technology we have is transformational, and if people aren't building their own systems they are going to miss where society is going.

But I have a counterpoint for myself

The inherent tension, as someone who works in media, is that I personally right now don't trust AI Agents fully. Why would I say that even after saying the above? Because as long as I'm sending something, anything, to a human, it must have a human mind to orient it to the other human mind. I kept talking about good, right? What to me is good is if a human being sent me something. The feeling of a real person shooting me an email, or someone personally overseeing a project and meticulously bringing it to life, is truly exciting for me. AI Agents and agentic systems are going to change a lot, but what we have to be careful about is that human beings, the same way we look for the label "organic" on a package, will reject the agentic systems we have and embrace technology and media that are cleanly directed by a human being. AI Agents are the future, and Humanity is the future as well.

As 2026 ends, we'll soon find ourselves saying that AI Agents finally and unquestionably succeeded. They are able to transform human engagement in truly powerful ways. What will make this good is if the human part is the element that succeeds.