Tilbury Spent $99 and Got $157. The Margins Are Embarrassing.
Mark Tilbury spent $99 over 7 days building three faceless AI influencers, Amos, Vivian, and Sienna, using Higgsfield and Claude. The winning character moved 23 sales of a $12 workbook, netting $157 profit. Not exactly a venture round, but the unit economics are at least transparent.
What this actually demonstrates is the concept of synthetic persona arbitrage. You create a fictional character with consistent visual identity, route its audience toward a single low-priced product, and the marginal cost of each additional video approaches zero. The mechanism is attention aggregation. The lesson, for those of you who failed basic economics, is that one hero product at sufficient view volume generates revenue without inventory, fulfillment, or your face.
Mark Tilbury, a creator who ran a structured 7-day experiment with a fixed $99 budget. He cloned existing AI influencer archetypes like Lil Miquela and Mother Satori, built three characters, and converted the best performer into a paid workbook.
- Open Higgsfield and generate a consistent AI character image. Save 5 to 10 variations of the same persona.
- Use Claude to write 7 short scripts for daily health-themed videos, each ending with a call to action about a gut-cleanse drink or similar product.
- Post one video per day for 7 days to a social platform. Track which post gets the most comments asking where to buy something. That is your market signal, not a vague feeling.