How AI plotted an interstellar journey to Alpha Centauri

Details

Physical Superintelligence (PSI), an AI physics research lab, launched today with $58 million from Breakthrough Energy — the climate-focused fund founded by Bill Gates. Its first public result: an AI-discovered trajectory that the nonprofit Fermi Explorer Mission will use to send a spacecraft toward Alpha Centauri, targeting a launch by end of 2029. The Fermi team spent a year failing to find a viable route for a $15M solar-powered spacecraft. PSI's 'Get Physics Done' system cracked the problem in about a week. The tool is open-source: it decomposes a physics research question into simulation subtasks and decides which to run, using AI models including Claude and GPT as reasoning layers. This is an agentic science workflow in production, not a benchmark claim. The mission itself is deliberately modest — the spacecraft carries at least 1kg of cargo including a copy of NASA's Golden Record, and travel time is up to 80,000 years. The goal contrasts sharply with Breakthrough Starshot: Yuri Milner's 2016 project pledged $100M to reach Alpha Centauri in 20 years using laser-propelled probes, and has yet to launch anything. Fermi prioritizes an actual launch over a distant promise. PSI's approach of AI-orchestrated physics simulations could generalize beyond space trajectories to other hard optimization problems — astrodynamics, materials science, climate modeling. The $58M launch and open-source tooling together make this a lab worth tracking.