About Chisomo Banzi

Chisomo Banzi began in architecture at Bond University before moving almost immediately into film production design, designing sets for major productions across Disney+, Apple TV+, Amazon, Netflix and NBCUniversal. After the 2023 Hollywood strikes disrupted the industry, he began looking internationally and eventually moved to Taipei in 2025, where he shifted his attention towards building spatial software. That has resulted in two very different products: Speakeasy3D, a VR conversation practice platform, and Serpentine3D, an open-source NURBS surface modeller developed as a free alternative to Rhino. Despite the change in medium, Chisomo sees a clear connection between them: whether designing a film set, VR experience or CAD tool, his work has always been about understanding three-dimensional space and how people interact with it.

Chisomo describes himself as largely self-taught rather than a traditional software engineer, although he has strong foundations in Python, Grasshopper, computer science, C++, Unreal Engine and spatial mathematics. AI coding became a necessity when he found himself trying to build ambitious software as a solo founder without the funding or team normally required. More importantly, it offered independence — allowing him to concentrate on what to build and why, while using AI to handle much of the implementation. His experience in production design has proved surprisingly relevant: designing a film set involves managing constraints, materials, budgets and spatial relationships simultaneously, and he sees software architecture as another form of systems thinking.

The benefits are speed and breadth, but Chisomo is equally clear about the limitations. AI works particularly well where iteration is fast, such as web development, but becomes less effective with Unreal Engine and VR, where visual Blueprints, recompiling, physics, spatial audio and performance optimisation demand greater hands-on understanding. AI-generated work also needs patience: sometimes the right solution is simply to discard what has been produced and start again. He also cautions against building a workflow that assumes today's inexpensive access to AI models will continue indefinitely.

His advice is to start with a problem you genuinely have and scope the solution ruthlessly. AI makes adding features deceptively easy, but every feature eventually needs maintaining. More importantly, don't attach your value to the code itself. As Chisomo puts it, the code may be rewritten by AI six months later; the lasting value lies in understanding the problem, designing the experience and building a community around the solution. Those are the things AI can't do for you.