The bioreactor is the experiment.

Why TL-BSB-7L is built around its boundary, not its contents.

Most bioreactors are designed around what is inside them. Volume. Mixing. Temperature control. The contents are the product, and the vessel is in service to the contents.

The TL-BSB-7L is built the other way around. The contents matter, but the boundary is the experiment. Every input is metered. Every output is captured. Every flux across the wall of the system is known.

That choice is not aesthetic. It is what makes the reactor a useful instrument for the kind of inference Terra Labs does.

What “boundary-instrumented” actually means

A closed bioreactor has a small number of physical pathways for mass and energy to cross its boundary. Gas in. Gas out. Liquid in. Liquid out. Heat in. Heat out. Light in.

On the TL-BSB-7L, each of those pathways has dedicated instrumentation.

• Gas exchange measured by inline mass flow and composition sensors on both inlet and outlet
• Liquid exchange measured by metered pumps and inline conductivity, pH, and dissolved gas sensors
• Thermal balance measured at the jacket and at multiple internal probe points
• Photon flux measured at the wall and corrected for absorption

The point is not to have a lot of sensors. The point is to have enough sensors, in the right places, that the boundary fluxes are not the unknown in any equation we write about the system.

When boundary fluxes are known, internal dynamics become the inferential target. That is the regime where hierarchical Bayesian methods do their best work.

Why 7 liters

The volume was not arbitrary.

A reactor needs to be big enough to support real biological dynamics. Microbial communities, photoautotroph cultures, and small invertebrate assemblages need room to behave like communities, not like isolated populations. Below about 5 liters, edge effects dominate. The wall is too much of the system.

It also needs to be small enough to instrument completely and run in parallel. A reactor that takes a full lab to operate is not a useful platform for the kind of replication that hierarchical inference needs. Hierarchical models get their power from variation across units. One reactor is a case study. Twelve reactors is a science.

7 liters is the smallest volume at which we could meet both constraints reliably. It supports community-scale biology, it can be fully instrumented at the boundary, and a single bench can run a meaningful number of units in parallel.

What the reactor is for

The TL-BSB-7L is not a production vessel. It is an inference platform. Specifically, it is designed for three classes of work.

Calibration. Closed bioreactors with known boundary conditions are the cleanest possible environment for fitting biogeochemical models. The priors and posteriors that come out of bioreactor work are what we carry into the field.

Hypothesis testing. Once a model fits the field, you can run perturbations on the reactor that you cannot run on a reef. Nutrient pulse, temperature shift, light regime change. The reactor tells you what the model predicts, and whether it does.

Hardware development. Sensors that will eventually go on the TL-04 rover or into a BLSS module are validated against the reactor first. If a sensor cannot resolve dynamics in a 7-liter bounded system, it will not resolve them in a 100-meter reef plot.

The connection to the field

A reef on the windward side of Oʻahu is not a bioreactor. The boundary is fuzzy. The fluxes are noisy. The volume is contested.

But the structural problem is the same. There is a bounded volume of water and substrate. There are inputs and outputs. There are internal compartments doing biogeochemistry. The inferential question is the same: given what we can observe at the boundary and inside, what is the state of the system?

The reactor lets us ask that question in an environment where we know the answer. The field lets us ask it in an environment where we need the answer. The same model fits both, because the system is the same kind of system.

This is what we mean when we say one framework, every bounded system. The framework is hierarchical Bayesian inference. The reactor is one of the cleanest implementations of a closed system we can build. The reef is one of the most consequential. They sit on the same continuum, and the TL-BSB-7L is the bench end of that continuum.

What comes next

The next post in this series will go inside the inference engine itself: model structure, prior specification, and how the hierarchy is organized across units, sites, and systems. After that, field notes from Hawaiian reef deployments.

Questions about the reactor, the program, or the methods are welcome at cameron@terralaboratories.com.

Terra Labs is a bioregenerative hardware and inference infrastructure company. We build the instruments, data systems, and models that make closed biological systems legible.