The number of designers and artists who use code is growing every day. One key advantage of the computational abstraction of ideas lies in defining rules and constraints to generate numerous artefacts from a single visual concept. In this context, the variability among the artefacts is often achieved in an exploratory step of manual parametric manipulations that affect the output. However, due to the number of parameters and the resulting combinatorial explosion, most designers are limited to exploring only portions of the design space. In this paper, we address this limitation by presenting the tool EvoProteus, which takes a Processing program as input and, using a Genetic Algorithm (GA), evolves parametric variations of that program. We also experiment with three methods to evaluate the visual outcomes produced by the evolved Processing programs: a human-guided, a prompt-based using CLIP and a hybrid strategy that combines the first two methods. In line with an emergent landscape of Artificial Intelligence (AI) applications to visual domains, we study the ability of the proposed tool to enhance the exploration of parameters in Generative Design and promote human-computer co-creation.