The Lab

Jaybird Labs is the research portfolio of James Thomas. The throughline across the work is a single question: how energy and dynamics shape the way intelligent systems evolve and act — studied across simulation, learning, and control.

Four explorations, one idea

Most projects here are the same question in different clothes. Four explorations orbit one central idea — energy & dynamics:

• Physical simulation — Hamiltonian systems, n-body dynamics, and the geometry of motion.
• Energy-based models — learning the landscape a system lives on, and sampling from it.
• Embodied control — robotics and sim-to-real transfer, where a policy meets mass, inertia, and noise.
• Language & reasoning — structured, energy-based views of NLP and inference.

The differentiator is interactive, playable work — research you can reach in and perturb, not just read about.

Sim-to-real: Visualize · Simulate · Validate

Within the embodied control work specifically, the cycle is Visualize → Simulate → Validate: make the math visible, stress-test it against synthetic reality, then judge simulation fidelity against the messiness of a real run. This is the signature of the embodied thread — not the whole lab.

Operator profile

James Thomas is an MS Artificial Intelligence candidate at the University of Texas at Austin, with a BA in Physics from Arizona State University. With over two decades of experience in software engineering, his research focus is on embodied agents, sim-to-real transfer, and the physics of intelligent behavior.

For long-form writing and research updates, visit jamesthomas.dev.