How do you work better with artificial systems?

I am practicing doing things more.

One of those things is writing.

Making this Blog

Creating this blog was accelerated with the use of a language model to write code. I used Anthropic’s Sonnet 4.5 with Claude Code.

I wanted:

• to use static Markdown pages
• to be able to write $\LaTeX$ easily
• the ability for people to comment

These requirements are pretty easy (though I’ve tabled the last bullet point because I did not like Giscus, the first recommendation from Claude). With some dedicated hours of free time and web surfing, I probably could have come up with something close to the initial website, as for something as common as a static blog, Claude Code’s suggestions are roughly the expectation value of standard Internet lore.

Why Haven’t I Blogged Before?

In the past I personally found the activation energy too high for me to build and maintain a blog.

However, with AI tools, the feedback loop connecting desiring-a-change to completing-a-prototype-fix is (for people like me) sped up by one to two orders of magnitude. What could have taken hours to figure out can take less than ten minutes. That boost in speed makes a qualitative difference when building.

The serious downside is that I will likely not know how to code projects like these with my own fingers. However, as a physics Ph.D. student by day, and someone who prefers to sleep by night, I was unlikely to dedicate time to understand how to build anything more than a minor software project anyways. For me, AI expands what I can do, leaving me more time to learn what I care about.

Developing Vibe Power

“Our supremacy on Caladan,” the Duke said, “depended on sea and air power. Here, we must develop something I choose to call desert power.”

Or consider this other quote from Dune by Frank Herbert:

“My father spoke of desert power,” Paul said. “The Harkonnens cannot rule this planet without it. They’ve never ruled this planet, nor shall they. Not even with ten thousand legions of Sardaukar.”

I speculate that one of the most important things in the nascent AI age is to develop your vibes. In the age of AI, “vibe power,” which correlates with the ability to leverage these systems to produce real results, will be critical.

Not everyone will be able to build cutting-edge AI systems. But with AI, everyone can build something new. By building out their individual desires, each person can strengthen their vibes by instantiating their dreams.

The alternative to building is to sit back and hope for a Butlerian Revolution to save us all. That won’t come.

Aligning to Vibes

Here’s why I think developing confident vibes is important.

Assuming that AI systems are well aligned (which is a significant assumption), there remains the question of to whom they align. We all can imagine the pessimistic scenario where they simply align to the highest bidder. One could argue this is roughly the current reality, where the rich—the Musks and Altmans and Zuckerbergs of the world—are able to craft AI in their image, or at least craft AI that serve their interests.

But there is a world we can build together: a meritocracy of ideas, where autonomous systems are moral, and work with us toward a collective good that broad swaths of humans can agree upon and flourish within.

We have to be bold enough to define the future we want and strive toward it together.

In the future I believe in, AI aligns with those with bright and articulate visions. It is the vision that will win, not the individual visionary.

Admittedly, this all sounds rosy and platitudinous even to my ears, and it might amount to copium that I am a meek graduate student instead of a rich tech entrepreneur. But a growing part of me believes it is possible.

Beginning to Build

The best investment I can make right now is to build ambitious deliverables. Build, document, and learn—not prematurely sell or publish or scale.

I need to see how far I can push these generative AI systems to figure out what I can and want to do. It’s not my job right now to advance these systems. Expansion of their abilities is inevitable. Someone, somewhere, will remove the limitations. The question is—then what?

In the widening sea of endless possibility, where will we swim?

In these intermediate months to years while artificial agents grow in capabilities but still lack the full richness of human vision and insight, I hope to leverage AI to create things I always found cool but never before could dedicate the time to do.

This blog is just the start.