This article is the first entry in a log of my first-ever open-source project. What I'm building is a concept called AI Taxiway: pooling idle phones, game consoles, and PCs so they can carry AI computation. I could take the step because AI drafts the documents an open-source project needs, such as the README and the license. And I'm going open because building alone would be lonely, and I want to find companions.
To be honest, I've never done open source before. I knew there was a world where people publish their code and grow it together with others. But I had never once stepped into that circle myself. So this is a complete first attempt.
Whether it will work out, I honestly don't know yet. I might give up partway through. So this is not a declaration that anything is finished. It's a record of how it feels at the moment of starting, written down just as it is.
What is AI Taxiway? A concept for pooling idle devices to carry AI computation
AI Taxiway is a concept for a system that pools phones, game consoles, and PCs that sit plugged in doing nothing, and has them carry AI computation.
All over the world, devices spend a lot of time plugged in and idle. Could that spare time be collected and woven into one large network of computation?
Right now, AI computing power is concentrated in a handful of huge places. Could a little of it be pulled back toward "everyone's hands"? That question is where AI Taxiway begins.
But let me say this plainly: I still don't know whether it can actually be done. People really are attempting similar things. Yet I don't know of a single example where a system like this has been properly established at large scale. My sense is that this is a hard problem nobody has solved yet.
The ambition is big, and the ground underfoot is uncertain. I want to try anyway, holding both at once. That's the stage I'm at.
Why could I start now? Because AI drafts the README and the license
There is one reason: AI is here. AI writes the first drafts of the whole set of documents an open-source project needs, so the heaviest part of the preparation is taken off my hands.
The me of eight years ago probably wouldn't have done this. Not because I didn't want to, but because the preparation you have to do before you can even start was too heavy.
Starting an open-source project means a mountain of work besides the code itself.
- A README that explains the project
- A license text that sets the terms of use
- A CONTRIBUTING guide for people who want to take part
- All kinds of other documentation
Unless you prepare this entire set, one piece at a time, the project never takes shape. For me, the real wall was never the idea. It was this comprehensive document-writing work.
That is exactly what AI fills in. If I say, "I want to start a project like this," it lays out a first draft of everything from the README to the license. It picks up the items I would likely miss. It takes over the work of filling in the missing documents, comprehensively.
What matters is this: AI writes the drafts, but I keep hold of the vision and the decisions. What to build, what to value, how much to hold and how much to open up. Those judgments stay with the human.
AI gets the documents in order, and I decide what kind of project this becomes. The moment I saw that division of labor was the moment I first thought, "Let's give it a try."
Why is deciding the rules of an open-source project fun?
Because it gives me the feeling that I am deciding the framework of this small world. It was an unexpected discovery I made once I actually started working. Deciding the rules is genuinely fun.
Open source is not just publishing your code. You have to decide, yourself, on what terms people may use it. Here is what I'm deciding right now.
- Make the license Apache 2.0, in a frame where commercial use is fine too.
- Decide how much I (the core) hold firmly, and from where I open things up to everyone.
- Don't lock in "this is the one correct answer." Adopt a philosophy where good things win out through competition.
- Properly record, in the documents, respect (credit) for the person who first proposed the idea.
- Set up rules for contributing (something like a DCO) so that people who take part can feel at ease.
I decide each of these, one by one, with my own hands. It's the fun of becoming a rule-maker. Before putting up any buildings, I'm writing the ground rules of the town. It turned out to be far more enjoyable work than I'd imagined.
What is my real motive? Not the AI boom, but wanting to make something amazing
In one line: it would make me happy to make something amazing. That's all. I'm not starting this because AI is trendy. Let me be honest here too.
"You're doing an AI project because AI is trendy." It might look that way. But I have a feeling the real motive isn't there. What's at the root is something much simpler.
It's less "a new technology arrived, so I'll ride it," and more a plain creative urge: there's an interesting wall, and it would feel good to get over it. The trend is, at most, a wind that happened to push me from behind.
I can't claim anything grand. A little modestly, but honestly: I want to make it, so I'm making it. That's the closest thing to my motive.
Why go open? Because building alone would be lonely
The most candid reason is that if I built it all alone, I'd probably be lonely. It isn't only about some fine cause. This is the most human part of the whole story.
The concept is big, yet when I picture myself assembling it silently, by myself, I get a little uneasy. Heading toward a hard problem alone, with no one to share the joy with, is honestly tough.
So I decided to open it up. Going open is also a choice not to be alone.
Somewhere, there may be someone who finds the same question fun, someone who will puzzle over it together with me. I want to find that person. Publishing the code means, for me, almost the same thing as going out to look for companions.
I'll keep recording from here
So, I'm starting my first open-source project ever. This is exactly the first entry in that log.
Whether it works out, I don't know. I might stop partway. Even so, I'll write down here, honestly, as far as I've come.
It's still in preparation, so when it reaches a form I can share, I'll let you know here again. If anyone reads this and thinks, "that sounds kind of interesting," I'd be glad if one day we could work on it together. Because alone, it's lonely.
FAQ
What is AI Taxiway?
A concept for pooling idle phones, game consoles, and PCs and having them carry AI computation. People are attempting similar things, but I don't know of any example established at large scale. Whether it can actually be done, I still don't know.
Why does having AI make it possible to start an open-source project?
Because AI writes a first draft of the whole set of documents: the README, the license text, and the CONTRIBUTING guide. It also picks up the items I'd likely miss. The decisions about what to build and what to value stay with me.
What license will it use?
Apache 2.0, in a frame where commercial use is fine. I'll also set up contribution rules, something like a DCO, so that people who take part can feel at ease.
When will it be released?
It's still in preparation. When it reaches a form I can share, I'll let you know here. If it sounds interesting to you, I'd be glad to work on it together one day.
