A Motorcycle for the Mind

When I wrote about building Spark in March, I mostly wrote about the architecture: persistent memory, model routing, coding sessions, and a scheduler that could run work in the background.

Those were the pieces I had to build, but they were never really the point.

The point was that I was tired of being the database administrator of my own life.

The Administrative Life

Modern life has been split into an absurd number of separate systems. Tasks go in one app. Notes go in another. Health data lives somewhere else. Books, finances, projects, messages, calendars, subscriptions, photos, saved links, and half-formed ideas all have their own homes.

Each tool is usually fine on its own. The problem is what happens between them.

You have a thought while walking, then need to decide whether it belongs in your notes, your task list, your journal, or nowhere at all. You remember something important from a conversation, but it is buried in a chat thread. You write something down on paper, then eventually have to transcribe it, categorize it, and remember why it mattered in the first place.

None of this is difficult work. That is what makes it so expensive. It takes just enough attention, over and over, that you can spend a day keeping your systems tidy without moving any of the things you actually care about forward.

I do not think most people need more productivity systems. I think many of us need fewer moments where we are forced to stop thinking and start filing.

The Tools Should Adapt

Traditional productivity tools assume that a person can cleanly separate life into categories before they have even had time to understand what they are thinking.

That has never been how my mind works. A thought about work can be connected to something I am reading, a conversation with Inwu, an idea for a project, or the vague feeling that I have been avoiding something important. Requiring me to decide the correct folder before I can capture it is not organization. It is a small tax on having the thought at all.

I wanted to be able to talk out loud, ramble through an idea, send a quick message through Telegram, or turn a paper note into something useful without first deciding what kind of thing it was.

That is the shape of the problem Spark is trying to solve. It is not meant to replace every app I use or become a giant dashboard that controls my life. It is meant to sit around the edges, hold enough context, and do the clerical work of connecting things that I would otherwise have to reconstruct manually.

A Motorcycle, Not a Driver

The title comes from a recent Naval Ravikant episode, where he extends Steve Jobs's old description of a computer as a bicycle for the mind. Naval's point was that AI may be a motorcycle for the mind, but "you still need someone to ride it, to drive it, to direct it, to hit the accelerator, and to hit the brake."¹

The analogy matters because a motorcycle does not merely get you to the same place with less effort. It gets you there faster, covers more ground, and makes a lot more possible. It also leaves less margin for error.

On a bicycle, a missed turn is usually just a missed turn. On a motorcycle, you can make a bigger mistake before you have time to notice it. You are moving quickly enough that attention becomes part of the machine.

I think AI works the same way. Spark can remember the context around an idea, surface something relevant from an old conversation, turn a rough request into a focused coding session, and keep background work moving while I am doing something else. It can reduce the friction between an intention and the work that follows.

But it can also help me move quickly in the wrong direction. It can give me ten reasonable options for almost anything, write convincing code, draft a plausible plan, summarize research, and make a bad idea look more organized than it really is.

That is why it should not be driving.

Judgment Is the Work

I wrote recently that taste is becoming the last moat. When execution gets cheaper, the scarce thing is not the ability to make something. It is knowing what deserves to be made.

The faster AI makes it possible to move, the more valuable that becomes. Speed is useful when you know where you are going. Without that, it mostly means you can miss more things and compound mistakes more efficiently.

I think the same is true in a more personal way. The valuable part of my day is not sorting information into its final category. It is noticing what feels important, connecting it to the rest of my life, and deciding what to do next.

Spark can handle more of the plumbing around those decisions. It can keep track of details I would otherwise lose, bring back context I forgot I had, and make it easier to turn a thought into an actual project. But the system only becomes useful when it stays connected to my values, my family, my interests, and my judgment.

That is also why I am not especially interested in an AI that does everything automatically. A system that makes decisions for me might save time in the short term, but it could also make me less involved in the choices that shape my life. I would rather build something that gives me more room to think clearly.

What I Want to Explore

Spark has become a good place to test this idea because I use it constantly. The rough edges are obvious when a system is part of your real life rather than a demo.

Over the next few posts, I want to go deeper into what I have built since March: how memory can be useful without becoming invasive, how I use coding sessions to build personal tools, why reliability matters more than impressive demos, and what it takes to let an AI system help without making it opaque or exhausting.

The technical parts matter, and I am proud of them. But I want to start here, with the problem underneath all of it.

I did not build Spark because I wanted another AI assistant. I built it because I wanted to spend less of my life managing systems and more of it deciding what is worth doing.

Footnotes

1. Naval Ravikant, “A Motorcycle for the Mind”, February 2026. ↩