It would be great if we could monitor anything in the body continuously. The continuous glucose monitor (CGM) was a great start, but there are so many interesting molecules to human health beyond glucose, but no one has successfully made monitors that work for other things. This is largely because display-based screening techniques for binders are detached from the ground truth physiologic data that matters. It would great to have a screening method that is coupled to a readout, like a GFP or nanopore, so that readout of a designed binder produces a number that is more meaningful. In a preprint with @mu_anand and other folks at Caltech, we ask whether existing architectures around protein binders tied to a GFP reporter can serve as this screening vehicle to speed up ML-design of new monitoring proteins. We took 18 OpuBC-cpGFP biosensors that our lab previously evolved toward nicotine, SSRIs, and other neural drugs, and screened them against 63 human health-relevant compounds, creating function data for 1134 sensor-ligand pairs. 124 pairs came back at a signal level we deemed suitable to begin directed evolution campaigns, with a little over half of those having readouts at a dynamic range that could make them immediately usable in application today. In total about a third of our compounds in our limited screen found a paired sensor worth evolving on. We ultimately think that PBP-cpGFPs could be useful to increase the hit rate of initial screens of possible continuous monitors. You can read it below.