STEEL EAGLE: TITAN STORM

AI Tools Experience

This game was developed by the editor over the course of three hours using two AI agents: Codex and Qoder. Codex ran on the GPT-5.5 model, while Qoder utilized Qwen3.7. I first used Codex GPT-5.5 to outline the concept of the game I wanted to build and had it formulate a development plan. After a few minor adjustments, I proceeded to execute the plan, and within less than 20 minutes, an initial model was generated. At first glance, however, I wasn't entirely satisfied — the overall visual design left much to be desired. For instance, the UI for weapons and equipment felt lackluster (which I later discovered online might be a known shortcoming of Codex GPT-5.5). There were also issues with battle mechanics and numerous combat bugs. Fortunately, after three or four rounds of prompt optimization, the gameplay issues were largely resolved. One drawback of Codex GPT-5.5 is that it does not support continuous development; my purchased tokens had to be consumed in separate sessions. During the five-to-six-hour breaks between sessions, I used Qoder to run the code. As it turned out, Qoder identified and fixed a substantial number of bugs — including pathfinding failures where tanks would get stuck — and, after being refined by Qoder Qwen3.7-MAX, the UI was significantly improved as well (see attached images). In summary, both AI tools have their respective strengths and weaknesses. Codex GPT-5.5 excels in rapid prototyping and speed, whereas Qoder Qwen3.7-MAX demonstrates superior capabilities in logical reasoning, bug fixing, and aesthetic refinement.