I've been leveraging a range of AI tools to assist with my current project, and their impact has been transformative. Many of my prompts require multiple iterations, but the iterative process itself has proven invaluable for helping me come up with a solution and just helping me better understand my problem.
In my experience, the quality of AI-generated results improves significantly when I provide detailed, context-specific information. This means being explicit about referencing sprite names and even pre-defined global variables. The more precise this input, the more relevant and accurate the output tends to be.
Additionally, I’ve found it beneficial to break down requests into smaller, more focused tasks. I usually try to implement features myself and only turn to AI when I encounter roadblocks -- particularly those involving intricate algorithms, mathematical challenges, or logic that requires a perspective that is new to me.
The AIs I've tried are especially effective when it comes to exploring different problem-solving strategies. Rather than seeking direct solutions, I often frame my queries around conceptual approaches, asking questions like, “I want to accomplish THIS -- what would be the most effective way to tackle this problem?” This not only yields practical advice but also helps me think more critically about my own design and development choices.