# RoboOI

> RoboOI is an early-stage research company. It tests whether living neuron cultures (organoid intelligence) can teach robots to adapt with fewer real-world trials, then ships what works as ordinary silicon software. Tagline: "Train with biology. Deploy on silicon."

## What RoboOI does
- A neuron culture on a multi-electrode array learns a robot task in closed loop (task state is encoded as stimulation; culture activity selects the action; feedback closes the loop).
- Pre- and post-training activity is recorded through a fixed encoder and decoder.
- The change is distilled into a recurrent "control prior" that initialises a conventional robot controller.
- The deployed robot runs on ordinary silicon. The neurons stay in the lab.

## What is and isn't proven (as of 2026-10-29)
- Proven by others: neuron cultures learn simple closed-loop tasks (DishBrain Pong, 2022; organoid Cartpole, 2026).
- Not yet shown: that a culture-trained prior makes a robot learn faster. RoboOI's Experiment 01 tests this.
- Experiment 01 design: 1 culture, 1 robot skill, 4 arms (trained, untrained, shuffled, from scratch), SAC and PPO baselines at matched budgets; primary metric: real interactions to 80% held-out success. Pre-registered; results will be published either way.

## Who it's for
- Robotics teams whose policies break under small shifts (camera pose, surface, gripper wear, part tolerance). A 3 cm wrist-camera shift can roughly halve a vision-language-action policy's success (arXiv 2603.05868).
- Labs working with multi-electrode arrays.
- Investors in the pre-seed transfer proof.

## Roadmap
- Months 0-2: single-culture transfer proof
- Months 2-4: measured sample-efficiency gain
- Months 4-6: replication across cultures and a second task
- Months 6-8: paid robotics pilot

## Key sources
- DeMarse et al. 2001, The Neurally Controlled Animat: https://people.biology.ucsd.edu/dwagenaa/abstracts/01-DWBP.html
- Kagan et al. 2022, DishBrain (Neuron): https://pmc.ncbi.nlm.nih.gov/articles/PMC9747182/
- Smirnova et al. 2023, Organoid intelligence (Frontiers in Science): https://doi.org/10.3389/fsci.2023.1017235
- Robbins et al. 2026, organoid Cartpole (Cell Reports): https://news.ucsc.edu/2026/02/brain-organoids-goal-directed-learning/
- AnyCamVLA, camera-shift robustness: https://arxiv.org/abs/2603.05868
- International Federation of Robotics, 2024 robot figures: https://ifr.org/ifr-press-releases/news/global-robot-demand-in-factories-doubles-over-10-years

## Contact
- Site: https://robooi.bio/