Created the entire Short Circuit visual identity using AI — including Shorty, the robot mascot — with a multi-tool approach that matched each AI to what it does best.
Short Circuit needed a visual identity: a mascot, a logo, brand assets. I started on Fiverr, looking for a designer who could create a robot character. My dad actually came up with the name "Shorty" for the mascot. The Fiverr results came back and the coloring was solid, but the character design wasn't what I had in mind.
So I tried something different. I took the Fiverr output to ChatGPT and asked it to generate a similar robot character with a different style. Option 1 became the Shorty you see across this entire site. From there, I iterated on poses — waving, holding a wrench, getting zapped by lightning — and used AI to build the brand system around the visual assets.
Used ChatGPT's image generation to create the Shorty mascot and full pose library. Started from the Fiverr reference image and iterated through multiple rounds: refining the character design, generating consistent poses (waving, wrench, zapped, sitting, laptop, confused), and establishing a cohesive visual style with the dark backgrounds and electric blue accents that define the brand.
Used AI to build the brand infrastructure around the visual assets: color palette definition, typography system, image integration into the website, asset naming conventions, and strategic placement across all project touchpoints. The key insight was using each AI for what it does best — ChatGPT for visual generation, AI for strategy and code.
Different AI tools have different strengths. ChatGPT crushed image generation while other AI tools were better for strategy and code. The best results came from knowing which tool to reach for — not from forcing one tool to do everything. That extends beyond AI too — starting on Fiverr gave me the reference point that made the AI output better. Sometimes the best workflow chains multiple tools together.
The structural decisions here were strong — Richie picked the right tool for each job, kept the scope tight (one mascot, defined poses, clean handoff to brand work), and loaded context well by bringing the full ChatGPT conversation and understanding the tool tradeoffs.
The gap: feedback precision and iteration instinct were reactive, not preventive. The transparent background issue was caught and fixed, but could've been caught in the generation phase with explicit specs in the prompts. Next level: bake the quality standard into the initial prompts instead of auditing after the fact.