HeyDonto AI Launches DFT Labs With a Peer-Reviewed Foundation and a Staged Plan to Test the Physics of Intelligence

The new research lab has published its foundational paper and begun testing the framework in public, the first two stages of a three-stage arc toward a foundational model built from the physics up

KNOXVILLE, Tenn.--(BUSINESS WIRE)--HeyDonto AI Technology today launched DFT Labs, a research company built around Data Field Theory, a physics-based framework for machine learning. Two of the program's milestones are already behind it: the foundational paper is published in the peer-reviewed journal Frontiers in Big Data, and the framework is now being tested in public benchmark competitions where independent judges keep score.

"The AI industry has settled on the idea that intelligence is something you buy with compute. DFT Labs exists to test whether it is something you can understand with physics," said Rivers Morrell, founder and CEO of HeyDonto AI. "We are doing it in the most unforgiving sequence we could design. Publish the mathematics in peer review, state plainly what does not work yet and then carry the framework into public arenas where the best labs in the world are the competition and independent judges keep score. Only after the framework earns its standing in public do we build our own model from the physics up. Every stage has to be won before we claim the next one. That discipline is what makes the ambition believable."

The Three-Stage Proof Arc

Modern AI mostly works by brute force: billions of parameters, enormous compute budgets and a black box that no one can fully open. Data Field Theory starts somewhere else. It treats learning as a field evolving on curved geometry, using the same mathematics physicists use for systems near a phase transition, and it asks a specific question: is intelligence a regime of organization that physics can describe?

DFT Labs is running that idea through three stages and labeling each one honestly: proven, in progress, or a target.

Stage one: The proof is done, and the mathematics is real. The foundational paper, "Data Field Theory: A Geometric Framework for Learning on Riemannian Manifolds With Synthetic Validation and Limitation Analysis," was authored by Reza Nehzati, founder and Chief Scientific Officer of DFT Labs. It establishes the framework's geometric foundation and validates four first-principles predictions on synthetic data where the geometry is known: phase-transition behavior as concepts form, a measured link between a spectral property of the trained system and its robustness on unfamiliar data, finite-speed propagation of information, and approximate rotational symmetry. On that synthetic ground, the model reached 89.2% accuracy, ahead of the standard baselines tested.

The same paper is just as clear about where the framework falls short today: on real data whose geometry isn't known in advance, like handwritten digits, performance drops sharply. The limitation is stated openly, together with the criteria that would refute the framework altogether.

Stage one claims only this: the physics holds up, the predictions came true, and the line between proven and not-yet-proven is drawn in ink.

The paper is also the first entry in a larger body of work: a foundational treatise, roughly 14 further manuscripts in preparation for the field's top venues, and a book-length treatment underway.

Stage two: Rather than making claims from its own laboratory, DFT Labs is now taking the framework into public, peer-structured benchmark competitions: fixed problems, independent scoring and fields of credible competitors. The framework is applied to existing models the team didn’t build.

The claim under test is precise: that Data Field Theory measurably improves models it is layered onto, on real-world problems, judged by people with no stake in the answer. The arenas were chosen deliberately to attack the paper's own published limitation: real data whose geometry does not arrive predefined.

First independent results are expected this fall, and DFT Labs will report them with the same discipline as the paper: what held, what did not and what would refute the framework.

Stage three: Begins only when stage two is won. Once the framework has been publicly confirmed to outperform competing frameworks, DFT Labs will develop it into a foundational model of its own: a physics of intelligence model, designed from Data Field Theory first principles rather than layered onto someone else's engine.

Because every learning step in such a system follows from stated physics, the model is auditable end to end and, according to the DFT Labs thesis, is dramatically cheaper to train than today's frontier models. DFT Labs states that stage three is a target, not a result and will not claim it until the model exists and is measured.

"We derived predictions from first principles and measured them. Four held, and we published the places the framework fails just as prominently as the places it succeeds," said Reza Nehzati, founder and Chief Scientific Officer of DFT Labs. "That published gap, real data with unknown geometry, is precisely the problem we are now testing in public, on ground we did not choose and cannot tilt. The deeper hypothesis, that intelligence is a regime of organization rather than a property of any one substrate, remains a prediction, and we have stated exactly what would refute it. The mathematics is the foundation. The science is the work of testing it in public."

The foundational paper is available at Frontiers in Big Data. Journalists and analysts may request the publication roadmap and program materials through the media contact below.

About DFT Labs

DFT Labs is the foundational research company of HeyDonto AI Technology. It develops Data Field Theory as an open, peer-reviewed scientific program, publishing limitation analysis and refutation criteria alongside every result, with a manuscript series and a book-length treatment of the framework in preparation. DFT Labs is headquartered in Knoxville, Tennessee.

About HeyDonto AI

HeyDonto AI Technology is a portfolio of AI companies built as one stack, with DFT Labs as its foundational research company. HeyDonto is headquartered in Knoxville, Tennessee.