FirstPrinciples
A non-profit research organisation founded in 2024, building Theo, an autonomous AI Physicist that reasons symbolically and validates hypotheses against experimental data.
FirstPrinciples is a non-profit research organisation founded in 2024 to build an autonomous AI Physicist, codenamed Theo. The system is designed to reason symbolically, formulate hypotheses and validate them against experimental data, with the stated aim of accelerating fundamental scientific discovery. Its moonshot objective is unifying quantum mechanics and general relativity by 2035.
The organisation employs 21 researchers, engineers and scientists working across artificial intelligence, machine and symbolic reasoning, theoretical physics, scientific hypothesis generation and validation, and large model training. Its specialised AI models exceed 8B parameters and have been trained on more than 2m scientific papers. Alongside Theo, the team has developed a suite of four tools intended to enforce scientific rigor in AI-driven research.
FirstPrinciples describes itself as fusing the urgency and efficiency of a startup with the openness and public purpose of a non-profit. The work is directed at fields including energy, materials and medicine, where fundamental discoveries are expected to have downstream applications. Further details are available at https://firstprinciples.org/.