Wielding the New Fire

Look around the room you're sitting in.

The glass. The metal. The plastics. The ceramics. The refined materials. The light. The heat. The ordinary objects that make modern life possible. Almost everything you can see exists because humans learned to control fire.

Fire was never just a tool. It was a core primitive. Once we could control it, everything downstream became possible. The societies that learned to wield it earliest and most carefully gained a compounding advantage.

Fire is the lens I use for AI: a primitive, not a tool. And unlike fire, it became widely accessible faster than we learned to use it well.

Writing, agriculture, electricity, and computing sit on the same shelf, and I'm not here to rank them. What fire gets right as a lens for AI is the shape of the thing: a capability that sits under other work rather than beside it. What's different this time is speed. Each of those earlier primitives took generations, or at least decades, to reach most hands, while AI is working its way into more layers of work, in more places, in just a few years. That leaves us holding a fundamental capability that most of the people who now have access to it still understand poorly. Like fire, it can destroy when used carelessly or with malice. Directed with skill, though, it can transform the work we already do, and it can create what was previously impossible or economically out of reach.

Most of the current conversation runs in a narrow band. Some fear replacement. Some are all in, spending hard on a capability they haven't taken the time to understand. Plenty more have decided it's a neat toy. The postures look nothing alike, and every one of them rests on a thin read of what's here. This capability is already changing things. The open question is whether we'll develop the judgment and craft to direct it well.

The neat-toy version is the most common, and the strangest.

That's the equivalent of discovering fire and using it mainly to keep animals away.

I've spent most of my life taking things apart to see how they work and what would make them better. That habit has always carried the same tax: time. Every teardown, every rebuild, every wrong guess costs hours you don't get back.

That's the part that changed. The loop between having an idea and knowing whether it holds up used to run in weeks. Now it runs in an afternoon. I'm not learning things I couldn't have learned before. I'm learning them fast enough to act on while they still matter.

That loop isn't hypothetical. When I first wrote this, a few weeks of it had produced three small tools: ollanet, FilePress, and IngotVault. Each went from itch to shipped site in days, faster than the old loop would have produced a plan. The weeks since have filled the shelf: a couple dozen tools, most of them public.

FilePress is a fair example of how that loop feels from the inside. The what fit in a paragraph: a folder of Markdown in git becomes a static site, with no admin screen and no content database, and the next build is the site. In the old loop, the first evening would have gone to build tooling and the second to parts I'd already solved on three earlier projects. Instead, a working version existed the same afternoon, which left my hours for the decisions the tool couldn't make for me, like what a draft means, what the content contract is, and what the build should refuse to publish. Those are exactly the hours that the mechanics used to crowd out, and several of our own product sites now run on the result, including FilePress's own.

Speed is the obvious gain. The quieter one is attention. Most of what filled my working hours was low-level load. Remembering the syntax. Holding six files in my head. Redoing the boring part of something I'd already solved once. Offloading that doesn't make me smarter. It hands back the attention I was spending on mechanics and lets me point it at the harder question of what's worth building at all.

Multiply that across a team, then across a year of compounding, and it stops being a productivity tip. It's a different operating capacity.

My bet is that the individuals and organizations that pull ahead will be the ones who treat this as a capability that rewards patient, deliberate practice, not a subscription to expense.

Fire didn't automatically produce civilization. People who learned to control it did.

The loop that used to take me weeks now takes an afternoon. The useful question is what you would point that at.