What the atlas is
The Chan Zuckerberg Biohub's single-cell initiative has now profiled tens of millions of cells across dozens of human tissue types. The output is a reference map: for each cell, a transcriptional fingerprint, a spatial coordinate, and increasingly, a history of what states it has passed through.
This is extraordinary infrastructure. But the scientific community is still largely treating it as a catalog rather than a computational substrate. That framing is the bottleneck.
The reframing
A cell atlas is a state machine description. Each cell type is a stable attractor in gene-expression space. The transitions between types — differentiation, reprogramming, response to injury — are the dynamics of that machine.
Michael Levin's group has shown that the state machine can be driven electrically, not just chemically. This is the critical bridge: if you can read and write cellular state via bioelectric signals, you have the inputs and outputs of a programmable system.
The Biohub atlas provides the state space. Levin's work provides the I/O layer. What's missing is the instruction set.
Implications
- Programmable morphogenesis is not a distant prospect — it's an engineering problem we now have the vocabulary to specify.
- The body's "software" is distributed across cell populations in a way that resembles massively parallel computation, not sequential instruction execution.
- Disease may be, in part, a computation error — cells stuck in the wrong attractor, or failing to transition correctly — which implies that reprogramming is as valid a therapeutic modality as gene editing or pharmacology.
What we're watching
Whether the next generation of Biohub publications starts to integrate bioelectric state alongside transcriptomic state. That integration would close a loop that has been open for a decade.