Since 2021, NASA’s Perseverance rover has achieved a number of historic milestones, including returning the first audio recordings from Mars. Today, almost five years later land on the red planethe has just achieved a new feat. Last December, Perseverance successfully completed a route through a section of Jezero Crater. traced by Anthropic’s Claude chatbotThis is the first time NASA has used a large language model to control the car-sized robot.
Between December 8 and 10, Perseverance traveled about 400 meters (about 437 yards) through a field of rocks on the Martian surface mapped by Claude. As you can imagine, using an AI model to plot a course for Perseverance wasn’t as simple as typing a single prompt.
As NASA explains, delivering Perseverance is not an easy task, even for a human. “Each move of the rover must be carefully planned, lest the machine slip, tip, spin its wheels or run aground,” NASA said. “So since the rover landed, its human operators have painstakingly plotted waypoints – they call it a ‘breadcrumb’ – for it to follow, using a combination of images taken from space and the rover’s onboard cameras.”
For Claude to complete the task, NASA had to first provide Claude CodeAnthropic’s programming agent, with the “years” of contextual data from the rover before the model can begin writing a route for Perseverance. Claude then proceeded with the mapping process methodically, stringing together waypoints from ten-meter segments that he would later critique and iterate on.
This is NASA we’re talking about, engineers at the agency’s Jet Propulsion Laboratory (JPL) made sure to double-check the model’s work before sending it to Perseverance. The JPL team analyzed Claude’s waypoints through a simulation they use daily to confirm the accuracy of commands sent to the rover. Ultimately, NASA says it only had to make “minor changes” to Claude’s route, one adjustment being because the team had access to ground-level imagery that Claude hadn’t seen during its planning process.
“Engineers estimate that using Claude in this way will cut route planning time in half and make trips more consistent,” NASA said. “Less time spent on tedious manual planning – and less time spent on training – allows rover operators to make even more trips, collect even more science data, and perform even more analyses. That means, in short, that we will learn a lot more about Mars.”
If the productivity gains offered by AI are often exaggeratedin the case of NASA, any tool that could allow its scientists to be more efficient will certainly be welcome. During the summer, the agency lost around 4,000 employees – which represents about 20 percent of its workforce – due to budget cuts by the Trump administration. By 2026, the president proposed reducing the agency’s scientific budget by almost half before the Congress finally rejected this plan in early January. Yet even if its funding is kept just below 2025 levels, the agency has a tough road ahead. He is asked to return to the Moon with less than half of the workforce this was the case at the height of the Apollo program.
For Anthropic, this is a major achievement. You may remember last spring Claude I couldn’t even beat Pokémon Red. In less than a year, the company’s models went from struggling to navigate a simple 8-bit Game Boy game to successfully charting a course for a rover on a distant planet. NASA is excited about the possibility of future collaborations, saying that “autonomous AI systems could help probes explore ever more distant parts of the solar system.”
