The sharing of Internet memes is increasingly popular form of expressing opinions and complex sentiments in an easily understood image. Marketers are exploiting the attention memes generate by enlisting meme influencers to create memes to promote products. I develop a machine learning algorithm that classifies meme images scraped from the meme aggregation sub-Reddit forums to generate a panel of meme posts. The data reveal time-series patterns in meme proliferation and quality, competition for attention across memes, learning-by-doing in meme creation, and potential mechanisms of the learning-by-doing. These findings suggest that the market for meme influencer will tend to be concentrated and may tend to become a superstar market.