I recently built a gradient image generator using GPT-5, and the process was surprisingly smooth. Here are some key takeaways from my experience.
Plan Before You Code
A crucial first step is to have a solid plan. GPT-5 includes a "thinking mode" that can generate a coding plan for you.
- How to use it: Simply prompt it with "think step by step" or manually switch to this mode on the ChatGPT website.
- Why it matters: This planning phase is essential, much like with other coding agents such as Davin or Opencode.
- Streamlined workflow: Previously, I used Gemini 2.5 Pro for planning and Claude 4 Sonnet for coding. GPT-5 now unifies this process, eliminating the friction of switching between models and maintaining a consistent context.
AI's Limitations in Interaction Design
While GPT-5 excels at coding, it still has limitations when it comes to abstract interaction design.
- The experiment: I provided GPT-5 with Justin Jay Wang’s excellent tutorial site and asked it to create and implement a plan for a gradient image generator.
- The result: The implementation was technically sound, but the final product lacked the vibrant, visually striking aesthetic I had envisioned. I was hoping for stronger layering and more refined color control.
- The takeaway: This experience suggests that in the age of AI, the role of skilled designers may become even more critical. They could evolve into "super individuals" or "one-person unicorns" who can bridge the gap between technical implementation and creative vision.