Summary
Vinson Weng, senior creative designer at AI Engineer, shares a candid firsthand account of how a single designer handled hundreds of design deliverables for the AI Engineer World’s Fair — a conference with 7,000 attendees, 140+ sponsors, 300+ speakers, and 600+ sessions. His AI design stack consists of Devin, ChatGPT, and Figma, with the workflow built around strict upfront design system definition.
The talk’s central argument is that AI tools produce consistent, on-brand output only when given a well-defined foundation: typography scales, color tokens, and atomic components established before any generation begins. With that foundation in place, Weng demonstrates automated pipelines including a speaker announcement generator that produces pixel-perfect branded graphics for all 300+ speakers from structured data, and a schedule export workflow where Devin pulls the latest session data and exports conference room schedules as print-ready PNGs — a task that previously required manual Figma work and extensive developer feedback loops.
Weng also addresses the limits of AI for vector asset creation, describing a practical workaround: prompting for raster images and vectorizing them in Figma rather than asking models to generate SVGs directly. The talk references Simon Wilson’s 2025 pelican-riding-a-bicycle LLM test as a framing device for thinking creatively about tool constraints. Practical and unpretentious, this is a useful case study for designers, design engineers, and small teams looking to dramatically increase output without additional headcount.
📺 Source: AI Engineer · Published September 10, 2026
🏷️ Format: Workflow Case Study
