AstroForge, a young startup focused on developing technologies for asteroid mining, is creating an autonomous control system for its spacecraft named "Solo." This system is based on a transformer model, a type of advanced artificial intelligence (AI) architecture known for its ability to process and understand complex data. The company plans to launch its first autonomous spacecraft in 2027 using the first rocket from Stock Space, a relatively new space launch company. This mission, supported by NASA, aims to collect scientific data about the Sun. While many spacecraft currently rely on traditional control algorithms for autonomy, concerns about the reliability of neural networks have limited their use in critical systems. However, last year marked the first time a neural network was used to control a satellite’s position in orbit. AstroForge was founded in 2022 and has raised $56 million in venture capital funding. The company has launched two prototype spacecraft, but both experienced technical problems that hindered their ability to complete their missions. In 2025, the company launched its Odin spacecraft into deep space, but it faced significant challenges in communication. Only a few large antennas on Earth are capable of transmitting signals to spacecraft located hundreds of thousands of miles away, and the opportunities to do so are limited. Despite these challenges, AstroForge was unable to regain control of Odin. This experience led the company to explore alternatives, such as embedding enough intelligence onboard spacecraft to allow them to solve their own problems. To achieve this, AstroForge’s co-founder and CEO, Matthew Gialich, decided to use the latest advances in transformer models developed by leading research labs. The company designed a system that combines traditional control algorithms with models trained on specific spacecraft subsystems, such as power generation and navigation. This system also includes an overarching intelligence layer trained on data from about 2,500 sensors aboard the spacecraft. Armand Awad, the company’s head of flight software, explained that the autonomous agent controlling the spacecraft will be able to perform tasks like identifying and resolving anomalies. For example, if the spacecraft loses track of its position, the system might detect a power issue related to its star tracker and fix it—possibly by simply turning the component on and off. AstroForge’s third spacecraft, DeepSpace-2, is scheduled to launch alongside Intuitive Machines’ third lunar mission, expected to occur by the end of 2026. Solo, the autonomous system, will be on board this mission, operating in "shadow mode," allowing engineers to test its capabilities before the planned Autonomy-1 mission. Gialich mentioned that for the Autonomy-1 mission, the spacecraft will not be equipped with radios that can receive signals from Earth, emphasizing its full autonomy. This approach represents a significant shift in how spacecraft are designed and operated, aiming to reduce reliance on ground control and enhance mission flexibility in deep space.