Organoid programming: definition, how it works and why it matters

Organoid programming is the practice of designing, running and reproducing experiments on living neural tissue as code. Stimulation protocols, input encoders, output decoders and closed-loop feedback rules are written, versioned and rerun like software, so that results from brain organoids can be compared across cultures, audited and built on by other teams.

Key takeaways

• Organoid programming is the engineering discipline underneath organoid intelligence: OI asks what living neurons can compute, organoid programming makes those experiments repeatable.
• You do not write instructions into neurons. You program the interface around them, and the tissue’s own synaptic plasticity does the adapting.
• A program typically defines four things: how inputs are encoded as stimulation, how responses are decoded, what feedback follows, and how everything is logged.
• Reproducibility is the central problem, because organoids vary between batches and change as they age.
• Open data formats and full provenance (cell line, culture history, protocol version) turn one-off demonstrations into cumulative science.

What does organoid programming mean? #

The term combines two ideas. An organoid is a three-dimensional culture grown from stem cells that self-organizes into tissue resembling part of an organ; a brain organoid develops neurons that connect and fire. Programming, in this context, means expressing an experiment as explicit, executable instructions rather than as a description in a methods section.

In practice, an organoid program is the code that sits between a researcher’s question and the living tissue. It decides which electrodes on a microelectrode array deliver pulses, how a value such as a game state or an audio feature becomes a spatial and temporal pattern of stimulation, how the network’s recorded action potentials are turned back into an output, and what happens next. When that loop runs continuously, it is closed-loop stimulation.

The phrase matters because it reframes the work. Landmark results in the field, such as cultured neurons improving at a Pong-like task (Kagan et al., 2022) and an organoid used as a reservoir for speech recognition (Cai et al., 2023), each depended on carefully engineered protocols. Writing those protocols as shareable, versioned programs is what lets other groups test, repeat and extend them.

How does organoid programming work? #

An organoid program usually has five layers. Each one can be developed and tested separately, which is one of the main advantages of treating experiments as code.

1. Culture selection. The program specifies which cultures are eligible: cell line, organoid age, and quality thresholds such as baseline firing rate or the presence of network bursts.
2. Encoding. An encoder maps inputs to stimulation: which electrodes fire, at what frequency, amplitude and timing. This is the “input layer” of neural encoding and decoding.
3. Recording and decoding. Electrical activity is recorded, spike-sorted and summarized, then a decoder turns it into an output such as an action or a class label.
4. Feedback. A rule decides what the tissue experiences next. In the DishBrain experiments, successful responses were followed by predictable stimulation and failures by unpredictable stimulation, a design motivated by the free energy principle.
5. Logging and export. Every stimulus, response and parameter is recorded with the protocol version and culture metadata, ideally in an open format such as Neurodata Without Borders.

What does an organoid program look like? #

In code, a simple closed-loop experiment reads like any control loop: reserve a culture, define an encoder and a decoder, then repeatedly stimulate, read, decide and give feedback. The difference is that the “model” being trained is not a set of weights in memory but the living network itself. The program never edits the tissue directly; it shapes the inputs and consequences the tissue experiences.

Organoid programming vs related terms #

Why does organoid programming matter? #

The field’s own roadmap for organoid intelligence (Smirnova et al., 2023) highlights the need for better interfaces, data infrastructure and reproducibility as organoids scale up. Those are software and process problems as much as biological ones.

• Reproducibility. Organoids differ between batches and change over weeks. Without protocol versions and culture history stored alongside raw data, it is hard to know whether a result reflects the method or the culture.
• Access. Building a culture pipeline and electrophysiology rig takes specialized staff and equipment. Remote platforms, such as the Neuroplatform described by Jordan et al. (2024), show that experiments can be run on organoids from elsewhere, which only works if experiments are expressed as programs.
• Comparison. When encoders, decoders and feedback rules are explicit code, teams can swap one component while holding the rest constant, which is how fields make cumulative progress.
• Accountability. Provenance and consent records attached to each run support the kind of embedded ethics that the Baltimore declaration (Hartung et al., 2023) calls for.

What are the limitations? #

Organoid programming does not solve the biological limits of today’s organoids. Current organoids are small, lack blood vessels (see organoid vascularization), and flat electrode arrays only sample the tissue that touches them. Programs also inherit the variability of the tissue: the same code can produce different behavior on two cultures, so good programs include controls, baselines and enough repetitions to separate signal from noise.

There is also a vocabulary risk. “Programming” can suggest more control than exists. Researchers do not program what neurons compute; they program the conditions under which networks adapt, and they measure what changes. Describing results in measurable terms, such as firing rates, response accuracy and the stability of learned behavior, keeps the term honest.

Where is organoid programming heading? #

As interfaces move toward 3D and high-density recording, and as organoids grow larger and live longer, the volume and complexity of experiments will rise. That makes shared tooling more valuable: protocol libraries, standard benchmarks, real-time runtimes and open datasets. For a broader view of the roadmap, read the future of biocomputing, and for the learning experiments that started it all, see how neurons learn in a dish.

Organoid programming is also the problem our company is named after. At Organoid Programming we are building the platform for it: protocols as code, a closed-loop runtime, and versioned, open data with provenance attached. If you are new to the field, start with our guide to what organoid intelligence is.

Frequently asked questions

What is organoid programming in simple terms?

Organoid programming is writing the instructions for an experiment on living brain tissue as code. The code says where and when to stimulate an organoid, how to turn inputs into stimulation patterns, how to read its electrical responses, and what feedback to give. Because it is code, the same experiment can be versioned, shared and rerun on other cultures.

Is organoid programming the same as organoid intelligence?

No. Organoid intelligence (OI) is the research field that studies whether brain organoids can perform computation and learning. Organoid programming is the practical layer underneath it: the protocols, software and data practices that let researchers run OI experiments reliably and reproduce each other’s results. You can think of OI as the science and organoid programming as the engineering discipline.

Do you program neurons the way you program a computer?

Not exactly. You cannot write instructions directly into neurons. Instead you program the interface around them: which electrodes deliver stimulation, how inputs are encoded, how responses are decoded, and how feedback is delivered. The tissue’s own synaptic plasticity then changes how it responds over time. The program shapes the environment, and the network adapts to it.

What tools are used for organoid programming?

A typical stack includes brain organoids grown from induced pluripotent stem cells, a microelectrode array for recording and stimulation, spike sorting and decoding software, a real-time closed-loop controller, and standardized data formats such as Neurodata Without Borders. Protocols are usually written in a general-purpose language such as Python.

Why does reproducibility matter so much in organoid programming?

Organoids vary from batch to batch, age over weeks and respond differently depending on culture conditions. If the exact protocol, culture history and raw data are not recorded together, other teams cannot tell whether a result reflects the method or the particular culture. Treating experiments as versioned code with full provenance is the most direct way to make results comparable.

Sources

1. Smirnova et al. (2023). Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Frontiers in Science.
2. Kagan et al. (2022). In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. Neuron.
3. Cai et al. (2023). Brain organoid reservoir computing for artificial intelligence. Nature Electronics.
4. Jordan et al. (2024). Open and remotely accessible Neuroplatform for research in wetware computing. Frontiers in Artificial Intelligence.
5. Rübel et al. (2022). The Neurodata Without Borders ecosystem for neurophysiological data science. eLife.
6. Hartung et al. (2023). The Baltimore declaration toward the exploration of organoid intelligence. Frontiers in Science.