Emergent spacetime
A premetric BF-type theory in which spacetime emerges instead of being assumed, checked numerically with CUDA.
- A cross-AI research pipeline: Claude proposes mathematical ansätze, Gemini synthesizes and plans, GPT critiques.
- CUDA verification for soliton mass and charge scaling, metric fixed-point stability, and Born-rule decoherence.
- The June 2026 report falsified the field-only version: without background support, the metric collapses to rank 1 in the void limit.
The question
Most of physics assumes spacetime and then puts things in it. I wanted to know whether a metric could emerge from something more basic: a premetric, BF-type field theory where geometry is an outcome rather than an input.
How the work is done
The research runs as a pipeline across models. Claude proposes mathematical ansätze. Gemini synthesizes results and plans the next step. GPT plays critic and tries to break whatever survived. Then the claims go to code.
Every claim gets a numerical check in CUDA. The suite currently covers soliton mass and charge scaling, stability of the metric fixed point, and Born-rule decoherence.
What failed
In June 2026 the falsification run came back against the field-only version of the theory. A metric built algebraically from a single covector and its exterior derivative is not guaranteed to be full-rank Lorentzian. In the void limit its rank drops to 1, and the geometry collapses. Condensate-only and tetrad constructions fail the same way.
What survived is narrower: a background-supported effective model that stays Lorentzian in every hybrid configuration tested.
The negative result is published in the repository alongside everything else. It cost me the version of the theory I liked most. It is also the most useful thing the project has produced so far.