- Organisation
- Neurorium
- Year
- 2026
- Substrate
- Biological neural networks
- Interface
- Electrophysiology + software
- Task
- Visual memory, classification and control demos
- Evidence
- Company demonstration
Neurorium is a Berlin biocomputing startup developing biological neural networks as experimental processing units. Its public demonstrations include short-term visual-memory reconstruction, image classification, few-shot learning and a simple steering-control task.
The company grew from work involving the Free University of Berlin and FinalSpark’s remote Neuroplatform. Its current material presents demonstration results without a peer-reviewed product benchmark, so this site labels the entry accordingly.
Why it is worth tracking
The demonstrations move beyond the now-familiar Pong benchmark towards vision and control. Those are relevant task families for a future biological coprocessor because they test temporal dynamics, adaptation and learned representations.
Evidence gap
For the database to upgrade this entry, useful next steps would include a detailed methods paper, fixed benchmark protocols, controls against conventional reservoir methods and repeatability across cultures.