TurPy: a physics-based and differentiable optical turbulence simulator for algorithmic development and system optimization

TurPy addresses a highly specialized niche in free-space optics (FSO) and astronomical imaging. Its primary value prop is 'differentiability'—allowing optical hardware parameters (like lens shapes or adaptive optics controllers) to be optimized alongside neural networks using standard backpropagation. While only 5 days old with 0 stars, the 6 forks indicate immediate interest from the academic community (likely peers of the authors). Its defensibility stems from the domain-specific physics implementation (subharmonic phase screens and autoregressive temporal modeling) which requires deep expertise in Kolmogorov turbulence theory and wave propagation. It competes with non-differentiable legacy tools like HCIPy or PROPER. Frontier labs are unlikely to enter this space directly as it is too niche for general-purpose LLM/Multimodal training, though it could be a component in specialized aerospace or satellite communication projects. The displacement risk is low because research-grade simulators often become 'sticky' within specific lab ecosystems once validated.