Through this three-week sprint, I’ve realized that A.I. tools operate as "Stochastic Parrots." They don't "understand" architecture in the way a human does; instead, they predict patterns based on vast datasets.
Pattern Recognition vs. Logic: While Midjourney can mimic the texture of brutalist concrete perfectly, it doesn't understand that a cantilever needs structural support.
The "Hallucination" Factor: In my text-based logs, I noticed the A.I. would confidently invent "facts" about building materials that sounded professional but were physically impossible.
The most significant takeaway from my exploration was the discrepancy between visual output and technical reality.
The Image AI Paradox: My "Spider-Web Skyscraper" prompt resulted in a stunning 8k render that could take a human weeks to draw. However, the plan was internally inconsistent—stairs led nowhere, and windows were placed where columns should be.
The Value of Research AI: Using Perplexity highlighted the importance of grounded data. While ChatGPT is a creative storyteller, Perplexity’s ability to cite real building codes proved that A.I. is only as good as the sources it can verify.