Astrohacker: A Super Intelligence Software Factory

This week I tried three words for what Astrohacker is.

First, a software factory: a system that turns issues into deployed software. Then a software garden: a living codebase that agents tend. Then, in A New American Super Intelligence Lab, a lab that applies super intelligence to the most important problems in the world.

Each word described something true. But only one of them describes the whole company. Astrohacker is a super intelligence software factory.

Why a factory

A lab makes something new once. A garden keeps a living thing healthy. A factory makes new things, again and again, and ships every one of them.

Look at what we build:

• EarthBucks is not a wallet app. It is a new electronic cash system, with its own blockchain, its own mining and, now, atomic swaps with Bitcoin Cash.
• TermSurf is not a terminal theme. It is a new terminal with its own protocol, which lets a browser, a chart or an image viewer run inside a terminal pane.
• NuTorch is not a shell config. It is a new shell with tensors built into the language.

None of these is an ordinary app. Each one is new twice: new as a product, and new as a technology. A lab could invent any one of them. What makes Astrohacker different is that the same system builds all of them, at the same time, and keeps shipping them.

That system is the factory. Every change at Astrohacker starts as an issue, and we solve each issue with experiments. Each experiment has a design, a result (Pass or Fail) and a written record of its evidence. Then the change ships. That is not a metaphor. It is how our factory works, every day, on every product.

The name has two meanings

Read “super intelligence software factory” one way, and it means a software factory that runs on super intelligence. Read it the other way, and it means a factory that makes super intelligence software. Both meanings are true, and that is why I like the name.

A factory that runs on super intelligence. This is what we do today. A team of super intelligent agents builds our products every day, from the design to the deployment. The agents write the code, run the experiments and record the results. People choose the problems and decide what ships.

A factory that builds super intelligence. This is where we are going. We will build our own models. The big labs build general models, and they are very good at it. But a general model is not the best tool for every problem. Electronic cash, cryptography and quantum computing all have their own data, their own rules and their own ways to check a result. A factory that works on those problems every day will learn what models they need, and then build them.

We already build some of the tools for that work. NuTorch, our shell, has tensors, modules and optimizers built into the language, so you can work with a model in the same place where you work with files.

The two meanings feed each other. The more we use super intelligence, the more we learn about what it cannot yet do. The more we build, the better our tools for using it get.

The lab and the garden are inside the factory

I do not want to throw away the lab or the garden. They are parts of the factory.

• The lab decides what to invent, and designs the experiments.
• The factory turns each decision into verified, deployed software. It gives every change a path, a limit and an end.
• The garden keeps the code healthy after it ships, so that the next experiment starts from good ground.

A lab without a factory writes papers and ships nothing. A factory without a garden ships once and then slows down. We need all three, and the factory is the one that turns the other two into products.

The factories we want to be like

The great factories did not only make things. They changed how things get made.

Ford’s moving assembly line at Highland Park cut the time to build a Model T from about twelve hours to about an hour and a half. Toyota let any worker stop the line when something went wrong, and made small improvements every day. Lockheed’s Skunk Works built aircraft that nobody else knew how to build, and built them fast. Each of them took a new way of working and used it to ship real things, again and again.

That is the model for Astrohacker, with one difference. Those factories had people. We have people and super intelligent agents, and the agents do not sleep.

What is true today

Astrohacker is in Austin, Texas. Our products are real, and you can use them today: EarthBucks, TermSurf, NuTorch and the rest of our products. Our agents build and ship them, with a record of every experiment.

We have not trained our own model yet. That is the next part of the factory, and we will write about it when we start.

So this is the answer to the question that I asked all week. Astrohacker is not only a lab and not only a garden. It is a super intelligence software factory: we use super intelligence to invent and ship new software and new technology, and we will build the super intelligence that the work needs.

If you want to work with a factory like this, as a researcher, an engineer, a partner or an investor, I want to hear from you. Get in touch.