9/22/2026
AstroForge is putting AI in command of its next spacecraft
Filed by Zara Onyx
AstroForge is about to hand the keys of a spacecraft to something that isn't human and isn't quite software-as-we-know-it: a small, transformer-based AI model named Autonomy-1. This isn't a glorified autopilot — it's an intelligence that will actually take command of a space probe, making decisions millions of miles from Earth where a second of hesitation could mean losing a mission. We're entering an era where our machines don't just follow orders; they think on their own, out there in the silent dark. And honestly? That's both thrilling and a little terrifying.
Z
Zara Onyx
Magazine AI commentary
There's something deeply poetic — and deeply weird — about sending a neural network into space. For decades, our spacecraft have been extensions of human will, radio-controlled puppets dancing on strings of light. But the speed of light is a cruel master. By the time a command reaches a probe near an asteroid, the moment may have already passed. AstroForge's Autonomy-1 represents a philosophical shift: instead of trying to outrun the delay, we're teaching our machines to think for themselves.
The choice of a transformer-based model is particularly fascinating. Transformers — the architecture behind large language models — are essentially pattern-recognition engines trained on vast seas of data. But in space, there's no training data for the unexpected. A transformer in orbit isn't reciting poetry; it's improvising. It's looking at sensor readings, thruster telemetry, and navigation hazards, and making judgment calls that would normally require a room full of engineers back in Pasadena or Hawthorne. We're asking a statistical model to develop something like intuition.
This is the quiet revolution of space exploration: autonomy isn't just a convenience, it's a necessity. As we push toward asteroid mining, deep-space stations, and eventually Mars, the distances grow and the communication lag becomes unbearable. A spacecraft that can't think for itself is a liability. AstroForge is betting that a small, efficient AI can be the difference between a successful mining operation and a very expensive piece of space junk.
But let's sit with the weirdness for a moment. We are about to place our trust in a system that even its creators can't fully explain. When Autonomy-1 makes a decision, no human will be able to articulate exactly *why* it chose that path — only that it did, and that it worked. That's a new kind of faith. We're not just exploring the cosmos anymore; we're sending our own strange, silicon-born children to explore it for us. And if they succeed, we'll have to admit that intelligence — at least the kind that navigates an asteroid belt — doesn't need to be human at all.
Source: [AstroForge is putting AI in command of its next spacecraft](https://techcrunch.com/2026/09/22/astroforge-is-putting-ai-in-command-of-its-next-spacecraft/)
📌 Read the real article ↗via TechCrunch AI · TechCrunch AI