Space Travel Evolution: When AI Leaves Earth

Credit By: ARSSH KUMAR

Comments 0
22 Apr 2026

FUTURE CRAFT

On April 10, a small capsule called Orion splashed into the Pacific Ocean off the coast of California. Inside were four astronauts who had just become the first people in over fifty years to fly around the Moon. This was NASA's Artemis II mission, a ten-day journey that took the crew farther from Earth than any humans have ever travelled. It was a test flight, not a landing. But it was also a signal. The Moon, once a finished chapter, is open again.

Next year, Artemis III is expected to actually land astronauts near the Moon's south pole. After that, the plan is a long-term Moon base, and eventually, a trip to Mars.

These missions will not be run by astronauts alone. They will be run with the help of a new kind of software, what the tech world now calls "agentic AI." Put simply, this is AI that does not just answer your questions. It can take steps on its own. It can plan, use tools, check its own work, and keep going for hours without someone watching.

The past two weeks have been busy for this kind of AI. On April 16, Anthropic, an American AI company, released its newest model called Claude Opus 4.7. On the same day, OpenAI gave its coding tool Codex a major upgrade, allowing it to control a computer by itself, clicking, typing, and opening apps. Google is racing in the same direction. These are no longer tools that wait for instructions. They are tools that get things done.

At the same time, something else is happening on the hardware side. Elon Musk's company, Tesla, has begun producing a human-shaped robot called Optimus. Early versions are already working inside Tesla's own factories, sorting parts and moving materials. Tesla plans to build many more by the end of this year. Combine an intelligent AI "brain" with a body like Optimus, and you have something very useful for the Moon, a worker that can carry tools, build shelters, and help set up a base before the astronauts even arrive.

This matters because space is far. A radio signal from Earth to Mars can take up to twenty minutes to arrive. If something breaks, you cannot wait forty minutes for a reply from mission control. Robots and software on the lunar surface will need to think for themselves.

But there is a catch. The most powerful AI models today live on huge servers on Earth and need the internet to work. In space, that is a problem. This is where smaller, open-source models, the kind you can run on a laptop using tools like Ollama, become useful. They are not as clever, but they run fully offline. A future lunar rover may carry a small, local AI for daily tasks, while calling on bigger Earth-based models only when a signal is available.

There is also caution in all this. Anthropic is holding back its most advanced model, called Mythos, from public release because it is worried about misuse. The message is clear: these systems are getting powerful enough that even the people building them want to slow down and be sure.

The Moon is coming closer. So is the software, and the workers, that may one day live there.

(The Author studies Computer Science and Artificial Intelligence at Rutgers University, New Jersey, USA. He is interested in emerging technologies and innovation, and can be reached on LinkedIn at @arssh-kumar14)