What I learned from this course is that AI in architecture is not really about replacing design thinking. It is more about testing how far a design idea can be pushed, and then learning when to stop, judge, and edit. Before this course, I mostly saw AI as a tool for generating images or making work faster. After using it through the assignments, I started to understand that AI can also become part of the design process itself: research, analysis, prompt writing, visual testing, comparison, and reflection.

One of the most valuable parts of the course was learning how to use different AI tools for different purposes. Perplexity was useful for research and precedent ideas. ChatGPT helped me organize thoughts, write prompts, and translate design intentions into clearer language. Image AI helped me visualize atmosphere, material, light, and spatial mood quickly. For my Floating Ground project, this helped me see how the concept could become more than a dramatic cantilever. It could become a climate-responsive student space with shade, ventilation, filtered daylight, and a stronger social ground.

At the same time, the course also showed me the limits of AI very clearly. When the task became long or too detailed, AI sometimes started to go off-track. It could forget earlier instructions, change the building too much, invent unrealistic structures, or make images that looked beautiful but did not make architectural sense. This was probably the most frustrating part. Sometimes the tool felt very smart at the beginning, then suddenly “went banana” when the task became more complex. That experience taught me not to trust AI blindly, especially in architecture where structure, site, scale, and human use matter.