For 28 years, every meter a Mars rover drove was planned by a human. An operator on Earth would study orbital images, identify safe terrain, plot a path, upload commands, and wait. The rover would execute the plan, send back results, and the cycle would repeat.
This week, NASA’s Perseverance completed the first Mars drives ever planned by an AI. Using Anthropic’s Claude vision-language models to analyze orbital imagery.
The human operators who have done this job since Sojourner in 1997 just watched a machine do it for the first time.
What Claude actually does
The AI analyzes high-resolution orbital imagery of the Martian surface. It identifies rocks, slopes, sand traps, and terrain hazards. It plots a safe driving path. It generates commands.
The same task a team of specialists would spend hours on, Claude handles in minutes. Not because the AI is smarter than the specialists. Because it can process visual information at a scale and speed that humans cannot match.
The specialists still review the plans. This is Mars. You do not send a $2.7 billion rover down a slope because a language model said it looked fine. But the planning workload, the tedious visual analysis of thousands of square meters of alien terrain, that is now delegated.
Why this is different from every other AI announcement
Most AI announcements are about chatbots, productivity tools, or benchmark scores. Things that matter to quarterly earnings and press releases.
This is an AI analyzing another planet’s terrain to help a robot drive safely across it. 225 million kilometers from the nearest human. With no margin for error. On a surface nobody has ever walked on.
The stakes are real. The application is meaningful. And nobody is doing this to optimize engagement metrics.
The 28-year transition
Think about what just happened from the perspective of someone who has spent their career planning rover drives. You developed expertise over decades. You know Mars terrain better than almost anyone alive. You have made judgment calls that kept billion-dollar machines safe on an alien surface.
And now a language model can do the first 90% of your job. Not because your expertise was worthless. Because the volume of data exceeded what human cognition can efficiently process.
This is the pattern that is coming for every field. Not replacement. Augmentation that eventually handles the routine, leaving humans for the edge cases and final decisions. The Mars rover specialists are not unemployed. They are reviewing AI-generated plans instead of creating them from scratch.
Whether that is a promotion or a demotion depends on your perspective.
What I cannot stop thinking about
Somewhere at JPL, an engineer who has planned Mars drives for 15 years watched Claude generate a plan in minutes that would have taken their team hours.
And that plan was good. Good enough to drive a rover on Mars.
That moment, the moment you watch a machine do your job competently for the first time, is the most defining experience in AI. Not because of what it means for the machine. Because of what it means for you.
Source: Crescendo AI