Hassan El Mghari identifies recurring signs of weak AI-generated interfaces, including predictable gradients, italic headings, decorative pills, excessive emoji and inconsistent spacing. Builders can turn those observations into explicit constraints and save their preferences in reusable agent instructions.
Hassan El Mghari recommends giving coding agents screenshots and visual referencesAn AI coding agent is a tool-using AI system that can inspect, modify, and validate software within a repository. instead of relying on vague style descriptions. Longer prompts that explain the intended users, interaction flow and desired lookPrompt engineering is the practice of designing and refining instructions, context, examples, and constraints to obtain useful AI outputs. produce stronger starting points, while splitting a complex build into focused steps makes each iteration easier to control.
Hassan El Mghari treats the first generated interface as a foundation rather than a finished product. Fast iteration with a capable modelIterative AI code generation repeatedly creates, runs, evaluates, and revises code instead of treating the first output as final., followed by deliberate refinement of layout, states, animation and brandingInterface refinement improves a functional user interface through repeated visual, interaction, state, accessibility, and brand-focused revisions., is what turns a functional draft into a coherent application.
Watch on YouTube
