100 synthetic Swiftie personas. 3 tools. One very specific obsession.
When Taylor Swift appeared on her now-fiancé Travis Kelce's podcast New Heights, it broke the internet — and the discourse that followed surprised me. Not every Swiftie was thrilled, and the range of fan reactions (plus the existence of the Gaylor theory) made me curious about what fandom really looks like under the surface. So I did the only reasonable thing: I built an app to simulate 100 Swifties and asked them directly.
Using ChatGPT's Deep Research, I generated a diverse dataset of 100 synthetic fan personas across 23 demographic categories — then built three interactive tools to explore the results. The essay digs into what happens when you try to model a fanbase with AI: how the data skewed too male at first, how 'happy' appeared 103 times across survey responses, and what the loudest voices on social media actually represent (spoiler: not much).
SwiftieGPT is what happens when you feed 100 synthetic personas into a Taylor Swift-themed survey, then build three interactive tools to explore the results. Blank Space Chat lets you talk to any of the 100 personas. The Eras Report visualizes their rankings. All Too Polled shows the raw survey data.
The strongest 'AI data product' in the portfolio — it combines persona generation, survey methodology, data visualization, and interactive chat in a single Replit app.

CustomGPT that generates 650-800 word sports cheat sheets using o3 + real-time Sports-Reference/ESPN data.

An AI-powered micro app that converts any table into styled charts with PNG/SVG export. Paste data, pick a palette, download instantly.

News headlines → dry New Yorker-style cartoon prompts → GPT Image 1 / Grok / Meta AI.