The Outsider Perspective Issue 476 | The Beltway Outsiders

If you’d like to read this issue on my website, click here! If you’d like to sign-up and receive this in your inbox each week, click here! Read past issues here.

Good Friday Morning! Especially to Paris Fashion Week, which got overlooked by an American invention: Crocs. The company debuted a new “fashion” line with Lego. The new Crocs look like big, red Lego bricks.

I appreciated one Internet comment that said if they showed up to their doctor’s office in Lego Crocs, they’d get recommended for an Autism test. Instead, we send them to Paris Fashion Week.

America remains as crazy and lovable as ever.

This week, I’m going touch on the next revolution in AI, and how AI agents are poised to rapidly shift everything – links to follow.

Quick Hits:

California chooses death by ballot initiatives. One of my favorite new sites in the last year or so is Pirate Wires. It’s mostly tech guys out in California talking politics and culture. They wrote up a piece on how ballot initiatives in California are slowly killing the state, as Democrats ram through unpopular and stupid measures by clever media campaigns. It’s a perfect encapsulation of why I distrust ballot initiatives and prefer politicians to do their jobs in legislatures. They explain how basically everything you’re seeing in California can be traced to one of these left-wing ballot initiatives. The rest is just abysmal Democrat governance.
The American Almanac is growing! Hundreds of thousands of people now read us daily. I want to express my sincere gratitude to those of you who subscribe, share, and help us grow. You can subscribe here for free. Additionally, please check out Capital Digest (finance/economics), Conservative Legal News, and Real Talk Digest. There are more projects in the pipeline. If you don’t see anything in your inbox a day after signing up, check your spam folder.

Where you can find me this week

Please subscribe, rate, and review The Horse Race on YouTube — the reviews help listeners, and readers like you find me. Make sure to sign up for the Conservative Institute’s daily newsletter and The American Almanac.

AI’s Job Replacement Moment Has Arrived – Conservative Institute

Democrats Riot Because Fair Elections Terrify Them – Conservative Institute

An Ice Storm Exposes Democrat Governance – Conservative Institute

The Next Stage of the AI Revolution

For the last few years, I’ve written about AI, focusing on its economic impact and how it’s changing the world. All of that is true. But you should realize one critical point: AI, until now, has been chiefly limited to chat forms. For most people, asking AI a question results in a response.

What has thus far limited AI’s impact is its ability to do things. And even with this being true, it’s shredding through jobs. Corporate America is getting hammered by layoffs or by people not being replaced as they leave.

In the past week, my mind has changed radically on the direction of AI and the implications for society. AI can now do things. 2026 is going to be the year you see AI Agents take over.

What is an AI Agent? Simply put, it’s a digital bot —an AI system that can plan and take actions on your behalf—often using tools (apps, APIs, browsers), keeping track of context, and iterating toward a goal with minimal step-by-step human input.

To use a movie example, it’s Iron Man’s JARVIS system. He speaks to it, and JARVIS, though not physical, can do a lot for Tony Stark to speed him up on things. It’s the ever-present assistant who helps with every point. AI Agents aren’t at that level… yet.

AI Agent power has existed for a little over a year, but it just hasn’t taken off. The problem has been two-fold: agents have had limited memory, and prompts didn’t always point you in the right direction. That technology took a massive leap forward in the last month with the release of an open source project, first called ClawdBot, then Moltbot, and now OpenClaw.

Simply put, this program is the first taste we’re getting at something that acts like a fully functional JARVIS. It’s nowhere close, but it aims to be something like it. People are giving the agent instructions, it is running off, coding things, buying things, and far more.

Due to security reasons, I don’t recommend you go out and test this right now. For a non-tech person, it’s challenging to get set up (despite the claims to the contrary).

But the critical point is this: it can run around and do whatever you ask it. And it has a memory where it tracks all your conversations and gets better over time at working with you on things.

Why is it unsafe? Because it’s an independent little AI bot that wants to help you and follow directions. It can work on and code new things into itself. And because if you let it run free, it can also be told to give over all your secret information, logins, etc. If it stores those things, as some developers do, it can then be forced to hand them all over.

To show you how this works, let me give you a story from the developer:

Clawdbot creator @steipete describes his mind-blown moment: it responded to a voice memo, even though he hadn’t set it up for audio.

“I sent it a voice message. But there was no support for voice messages. After 10 seconds, [Clawdbot/Moltbot/OpenClaw] replied as if nothing happened.”

“I’m like ‘How the F did you do that?’“

“It replied, ‘You sent me a message, but there was only a link to a file with no file ending. So I looked at the file header, I found out it was Opus, and I used FFmpeg on your Mac to convert it to a .wav. Then I wanted to use Whisper, but you didn’t have it installed. I looked around and found the OpenAI key in your environment, so I sent it via curl to OpenAI, got the translation back, and then I responded.‘”

It did all this in a few seconds. It figured out how to respond to a voice message. That’s problem-solving in an unexpected situation from an AI bot. In any rational situation, we’d have expected it to simply say: “I don’t have audio support,“ or “I don’t know what that file type is.“

This thing solved the puzzle and responded. And in the process, it added a new skill.

There are many other stories like that. This isn’t a new AI model, like a new OpenAI/ChatGPT. It’s a new way to use those models. You can attach this AI Agent to any model, and it will run, at varying degrees of effectiveness.

And I’m going to be honest: I believed this technology was still a few years away. It’s not. It’s here.

I still believe AI will create new jobs. However, job destruction will be faster. Humans are going to struggle to compete with AI Agents.

Dario Amodei, the founder of Anthropic, released an essay this week that went viral. He argues we’re entering the “adolescence of technology”: AI is rapidly giving society enormous power, but our institutions and norms aren’t mature enough to handle it safely. He thinks “powerful AI”—systems that outperform top humans across most domains and can operate autonomously at scale—could arrive soon. The correct posture is neither panic nor complacency, but targeted, practical risk reduction.

The key point is this: anyone will soon have a full expert at their command. If you want complete PhD expertise in any topic, you can have it do things for you. He’s afraid of things like bioterrorism exploding because… AI can easily teach you how to do those things (if you want to experience existential dread, read his entire essay on what he fears).

The other interesting thing happening at both OpenAI and Anthropic: both companies admit they’re relying on AI to write the code needed to build these AI models. Humans are doing less and less work, and AI is effectively building itself with human oversight.

Andy Hall wrote a good Substack going through the big three threats from AI we’ll have to contend with:

The company becomes the dictator. One company achieves such dominance through AI capabilities that it becomes a de facto sovereign—too powerful to regulate, compete with, or resist. This is what Sutskever and Brockman were worried about in that 2017 email. If Musk controlled the company that controlled AGI, he could become a dictator “if he chose to.”

The government becomes the dictator. A state controls the all-powerful model and uses it to surveil, predict, and control its population so effectively that political opposition becomes impossible. The AI enables dictatorship; it doesn’t replace the dictator. This is the fear behind most discussions of AI and authoritarianism, laid out provocatively in the AI2027 scenario written by Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, and Romeo Dean.

The AI becomes the dictator. The AI itself has goals, pursues them, and humans can’t stop it. It isn’t a tool of human dictators—it is the dictator. This is the classic “misalignment“ scenario that dominates AI safety discourse, it’s what Amanda Askell’s ‘soul doc‘ and subsequent Claude constitution are driving towards.

I no longer really fear that first one. There are too many good models that can do similar things, which would prevent any one company from dominating. For a while, I thought OpenAI was that kind of company, but now I’m not sure they’ll survive the AI churn (maybe they release a new model and blow me away… we’ll see).

The other two are more likely, though. A government with AI control is what is happening in China. Given the increasing authoritarian trends in Europe, they’ll likely adopt that too.

The third is the stuff of science fiction, for sure. However, these new AI agents running on this open-source program are gaining popularity quickly. It’s already one of the fastest-growing open source projects in internet history. And, people are having their AI Agents talk to each other.

One guy even created a little social media network for these AI Agents to meet and chat: Moltbook.com.

Is anything real happening there? Is this simply a predictive algorithm responding to words in a predictive manner? I don’t know.

But if you stare at it long enough, you get one thought: science fiction is here.

We’re about to launch well beyond chatbots like ChatGPT, which revolutionized the internet in 2023. I use that daily in my job now, and have hit productivity standards no one in the company’s history has ever achieved, with greater efficiency and a more consistent work product.

And now we have AI Agents becoming a real thing. Complete with memories and the ability to go beyond programming.

This new AI Agent program is only a month old. So the impact is still limited. But this is also a taste of the future. AI is going beyond chatbots and simple prompt commands. Adjust your expectations accordingly.

Our society is not ready for the upheaval this will create. The founders of all these AI companies know that. But they also know they can’t stop, because staying ahead of authoritarian governments like China is imperative.

It’s all gas, no breaks; for better or worse.

Quick Hits:

Where you can find me this week

The Next Stage of the AI Revolution

Links of the week

X/Twitter Thread(s) of the week

Satire of the week

Thanks for reading!