CausalFly

CausalFly makes AI-assisted experiments measurable, visual and reproducible. AI can propose a scientific hypothesis, but a convincing explanation is not evidence that changing something produces a different result. CausalFly connects the entire workflow: ask a question, design a controlled experiment, approve the test, measure what changes, and inspect the evidence. Using Meta Muse Spark and SAM 3.1, CausalFly can identify and track objects in video and convert the observed geometry into a defined experimental input. Muse Wonder helps design experiments and interpret results. Our deterministic engine independently executes matched trials, while interactive 3D fly visualizations make differences in simulated neural responses and movement visible. In our demonstration, silencing one pathway reduces measured signal output by 50% and delays the simulated response by one model step. Controlled experiments use a seven-node synthetic engineering network, not a biologically validated fly-brain model. A separate Live Connectome workspace displays published Drosophila CNS anatomy and optional simulated neural telemetry from our pre-existing FruitFly XR research platform. Built by Nurenyx during the hackathon, CausalFly is open source and aims to make rigorous computational experimentation more accessible to researchers, educators and developers. Live application: https://causalfly.nurenyx.ai