Detailed Analysis
A solo 3D artist with no formal coding background has spent 10 months building "Koryos: Sons of Mars," a large-scale ancient warfare RTS in Unreal Engine 5, using Claude as their primary programming collaborator. The developer, who specializes in high-to-low poly modeling rather than software engineering, has now launched a public playtest on itch.io. What makes this case notable is not the existence of the game itself but the methodology: rather than relying on complex prompt engineering, elaborate system prompts, or pre-built frameworks, the developer used a deliberately simple, iterative approach—building basic soldier movement first, then layering in naming and stats, then combat abilities, then enemy AI, with each stage building directly on the last.
The workflow described here is instructive for understanding how non-technical creators are actually using Claude for substantial software projects, as opposed to how AI-assisted coding is often discussed in more technical circles. The developer explicitly contrasts their approach with "complex coding setups," emphasizing instead that success came from extreme specificity in descriptions and patience when Claude produced incorrect or confused output. Rather than treating errors as failures requiring a different strategy, the developer simply continued iterating, testing, and correcting—treating Claude less like an autonomous agent and more like a patient, instructable collaborator that required precise communication to be effective. This mirrors a broader pattern seen among non-programmers using large language models: success often hinges less on sophisticated prompting techniques and more on breaking large, ambiguous goals into small, verifiable increments.
This case also illustrates a meaningful shift in who can realistically attempt game development, historically one of the more technically demanding creative disciplines due to its requirement of blending art, design, and systems programming. A 3D artist producing a functioning RTS with unit commanding, stat systems, combat mechanics, and enemy AI represents a genuine expansion of who can execute on ambitious software projects. Unreal Engine 5's C++ and Blueprint systems have a steep learning curve even for trained engineers, so the fact that a non-coder navigated this over many months using Claude as a translation layer between creative intent and executable code points to AI coding assistants lowering the barrier to entry for complex, multi-system software rather than just simple scripts or prototypes.
More broadly, this fits into a growing trend of solo and small-team developers using AI models to compress development timelines and skill requirements that previously necessitated hiring specialized engineers or years of self-study. It also reflects an important nuance in the AI coding assistant discourse: much of the public conversation focuses on professional developers using agentic coding tools for velocity gains, while cases like this show a parallel population of hobbyists and career-changers using the same underlying models for capability gains—doing things they simply could not do before, rather than doing familiar things faster. As tools like Claude continue to be embedded in creative and technical workflows, stories like Koryos: Sons of Mars serve as real-world evidence of AI's potential to democratize access to technically demanding creative fields, provided users adopt patient, incremental, highly specific interaction patterns rather than expecting one-shot solutions to complex systems.
Read original article →