Mission
Breakthroughs in our understanding of nature have long driven progress across society. Our mission is to understand the nature of reality by advancing AI-driven discovery in fundamental science.
Bottleneck
For as long as humans have existed, we’ve always tried to understand the world around us. What we’ve learned so far is remarkable, but with progress comes increased complexity.
The growing burden of knowledge is slowing scientific progress, because human cognition can’t scale with the proliferation of information available.
Shift
The scientific method is also evolving. AI is moving from a tool to a research partner, creating a fundamental shift in how (and where) science is performed.
Academic systems are built on human-led discovery, whereas frontier AI is concentrated primarily within labs with commercial priorities. For AI to meaningfully participate in the scientific process, how it’s built (and who it’s for) matters deeply.
Stewardship
As AI becomes foundational to discovery, it must protect knowledge as a shared asset, not a proprietary advantage.
FirstPrinciples is building AI solely for fundamental research, transparent by design, and committed to knowledge as a public good.
Theo: Domain-Specialized AI for Science
A modular system built to reason like a scientist as infrastructure for AI-enabled discovery in fundamental science.
Core design
Specialized fine-tuned models across physics domains; 8B+ parameters; trained on curated scientific corpus.
Key capabilities
Hypothesis generation, symbolic reasoning, tool integration, validation loops, Dynamic Research Objects (DROs) for reproducible steps.
Progress highlights
Early self-guidance achieved to move from question → hypothesis → evaluation; 10+ specialized models; internal experiments running on quantum information questions.
How it's different
Model-layer innovation for rigorous, interpretable reasoning; built with the scientific community to ensure alignment with the scientific process.
10+
Fine-tuned specialized models
8B-1T+
Parameters
Our Core Pillars
Scientific advancement
We enable new modes of discovery by building AI systems that contribute to rigorous, original scientific research.
Shared understanding
We aim to deepen humanity’s understanding of the universe and contribute to shared knowledge that supports long-term progress.
Responsible AI infrastructure
We work closely with the global scientific community to build AI infrastructure that can contribute meaningfully to scientific progress.
Institutional innovation
We are an independent, global company building AI systems designed for scientific discovery and grounded in the standards of rigorous research.
Explorations in Science and AI
Read more
Conjecturing
AI system 'Theo Conjecture' solves 35-year-old math conjecture, finds a term no one predicted
An automated discovery system built at FirstPrinciples by Randy Davila has proven a 1989 prediction by Paul Erdős and William Staton linking prime numbers to the Riemann zeta function, and uncovered a second mathematical term that had gone unnoticed for decades.
AI in science
Lab Notes: CiteVerify - Fake citation detection and evidence verification for AI research
Scientific writing has always relied on the implicit contract that claims must remain traceable to evidence. Large language models complicate that relationship. Modern systems can generate reports that appear rigorous and extensively sourced while quietly introducing fabricated citations. CiteVerify is FirstPrinciples’ answer to the rise of fake citations.
Conjecturing
Automated conjecturing: How machines are exploring mathematical structure
Before a theorem can be proved, someone has to decide what is worth proving. For decades, the formation of a conjecture remained largely outside the reach of machines. In Automated Conjecturing with TxGraffiti, Randy Davila explores how that boundary has shifted.
.avif)
AI in science
Between tools and theory: Reflections from the Machine Learning and the Physical Sciences workshop, NeurIPS 2025
A reflection on ML4PS 2025, where researchers in physics and machine learning grappled with the role AI should play in scientific discovery, and what it would take to move from process acceleration toward deeper scientific insight.