Skip to content
← /writing
May 23, 2026 · 6 min read · ai-agents · vibe-coding · agency · democratization

Fire for the Rest of Us

Vibe coding gave developers AI leverage. I took the same principles out of the editor and into my own life and career, and the lessons transfer.

In the myth, fire belonged to the gods. Prometheus stole it and gave it to people, and everything downstream of warmth, cooking, metal, and light followed from that one act. The point of the story was never the flame. It was who got to hold it.

I keep coming back to that myth, because AI is the most fire-like thing I have witnessed in my lifetime. I am not Prometheus, and this is not a humblebrag dressed up as mythology. I mean something specific: a capability that used to belong to a small priesthood is now in ordinary hands, and the people holding it are about to do things no one planned for.

## The fire is agency

For most of computing history, turning an idea into working software required a guild. You needed years of training, or the money to hire people who had it. The idea was never the bottleneck. The bottleneck was the distance between having an idea and being allowed to build it.

That distance is collapsing. With an agent, someone who has never written a line of code can describe what they want and watch it get built. Not a toy. A real thing that runs.

Speed is the obvious part. The real shift is agency: the permission to try.

## What the coders figured out first

Developers got there first, because code gives the fastest feedback loop on Earth. In early 2025 Andrej Karpathy named the new way of working "vibe coding": describe what you want in plain language, let the agent write it, steer by results. Within a year it was the Collins Dictionary word of the year.

Then Gene Kim and Steve Yegge wrote the book on doing it well, and gave the payoff an acronym: FAAFO. It is a wink at the old "find out the hard way" meme, turned into something useful. It stands for Fast, Ambitious, Autonomous, Fun, Optionality. You move faster. You attempt bigger things. You do alone what used to take a team. You actually enjoy it. And instead of betting everything on one approach, you try five and keep the one that works.

When the cost of trying drops to near zero, the question stops being "are you allowed to build this?" and becomes "what do you actually want to make?"

Read that list again. None of it is really about code. It is a description of what changes when a tireless, capable partner sits next to you and trying something stops being expensive.

## I took it out of the editor

That is the moment it clicked for me. I get FAAFO at work every day: I help build AI at a defense company whose mission is the world's best AI pilot, an agent that flies real aircraft.

I was an ok coder, nothing special, and that is exactly the point. Running this loop, I have built websites, stood up language models from a box under my desk all the way to frontier scale, and learned what it takes to run them reliably enough to be trusted in the highest-stakes national-security work there is. The same loop got me there: I used it to accelerate my education, to break into selling AI and machine-learning capabilities to the government alongside the engineers building them, and, honestly, to help land the job itself.

Then I pointed the same loop at the ordinary stuff: what I wanted to learn next, the household paperwork, the money questions, the decisions I kept putting off. Give it real context, set clear boundaries, check its work, keep what is true, drop the rest. It is the same discipline whether the agent is flying an aircraft or sorting your inbox. The lessons transfer.

It works. Not magically, and never without the craft, but the leverage developers have been quietly enjoying turns out to be available to anyone willing to run the loop. Almost no one outside the coder world has been shown how.

Learning is where I felt it most. I have explored and built things I dreamed about for years, because for the first time I had an infinitely patient teacher: one that explains anything, as many times as I need, without ever making me feel slow. It is the most powerful knowledge-amplifying technology of my life, not because it knows everything, but because it meets you where you are and walks forward from there.

My own career and life bent for the better around this. That is not a pitch. It is just what happened, and it is why I cannot stay quiet about it.

## Fire burns

The myth gets one thing right that the hype skips over: fire is dangerous. The same flame that cooks dinner burns the house down. The same agent that drafts your email can confidently do the wrong thing, leak what it should not, or get steered by someone trying to fool the person holding it.

So the leverage comes with a craft, and most of that craft is safety: keep a human on anything that cannot be undone, never let it touch your money or passwords on its own, and treat every urgent demand as something to verify first. That discipline is the part of my day job that matters most here, and it transfers cleanly: the people who benefit most from this fire are the ones who learn to handle it. Handing someone fire without teaching them to hold it is not generosity. It is negligence.

## Why I am writing this down

I cannot be the only one saying this out loud. The story we keep telling about AI is either doom or productivity charts. Both skip the human thing in the middle: a kind of power that was locked in a temple is loose in the world, and ordinary people can pick it up.

The playbook the coders found is learnable, and it was never really about coding. I went to the frontier for my job, I have spent a long while translating that playbook out of code and into ordinary life, and the most useful thing I can do now is hand it down, carefully, to people who were never supposed to get near it.

The fire is already out of the temple. The only real question is who gets to hold it.

I think the answer is everyone.


Part of the reliable AI-delivery trail. Browse the curated paths or inspect the proof.