Blog
Insights, tutorials, and updates from the Superset team.
Deriving Maxwell, Einstein, and Yang-Mills from Symmetry Alone
Our system derived the fundamental field equations of electromagnetism, general relativity, and the strong nuclear force from nothing but a field and its symmetry. No equations assumed. No Lagrangians provided.
One Equation Governs Earthquakes, Blood Flow, and Bridge Failures
A single three-parameter equation unifies regime transitions across ten physical systems, validated on 193 published measurements from 12 independent research groups spanning 1918 to 2013.
Our ML System Just Independently Derived the Alternative to Dark Matter From Raw Telescope Data
Given only 42 black hole measurements, our system mathematically proved that a hidden acceleration constant must exist, independently rediscovering the foundation of MOND.
New Discoveries across Five different subfields of Physics
From cosmological distances to turbulent boundary layers, we have discovered closed-form formulas spanning cosmology, general relativity, nuclear and stellar physics, fluid dynamics, and radiative processes.
A 4-Parameter Nuclear Charge Radius Formula That Autonomously Derives the Neutron Skin Effect
A simple algebraic formula replaces decades of supercomputer simulations, predicting the size of 952 atomic nuclei with 0.023 fm accuracy and autonomously discovering the neutron skin effect.
Recovering Einstein's 1924 Condensation Law and a Hidden Quantum Identity From Raw Numbers
Our system exactly recovered the Bose-Einstein condensate fraction law and independently discovered a hyperbolic identity in quantum tunneling, both from numerical data alone.
2,797 Pulsars Reveal That Spin-Down Depends on the Interstellar Medium
Analysis of the ATNF Pulsar Catalogue shows that pulsar spin-down rate correlates with the surrounding interstellar medium density, a relationship not predicted by the standard magnetic dipole model.
A 3-Term Formula for Wind Turbine Blade Noise From 1,503 NASA Wind Tunnel Tests
Analysis of NASA airfoil experiments reveals that aerodynamic noise can be predicted from three simple terms, replacing hours of computational fluid dynamics with instant calculation.
Beyond Estimation: Achieving 3.67-Cycle Precision on NASA's Lithium-Ion Benchmarks
We validated our battery degradation model against NASA's B0005 dataset, achieving 92.5% accuracy with an average error of just 3.67 cycles on completely blind test data.
Beyond Consensus: Finding Macro Alpha with High-Precision Retail Sales Forecasts
How our ML engine delivers a 0.38% Mean Error on US Retail Sales, giving traders a critical edge before the print.
The 90-Day Edge: Trading Energy and Agriculture with Institutional-Grade Weather Intelligence
How our ML engine leverages ERA5 data to predict temperatures 3 months out with 1.51°C precision.