what's /backflip?

/backflip is a generative NFT collection by Nikita Diakur consisting of 56 video artworks extracted from the training process of a digital avatar learning to perform a backflip. Using machine learning techniques from the DeepMimic paper ("Example-Guided Deep Reinforcement Learning of Physics-Based Character Skills"), the avatar—a self-portrait of the artist—iteratively imitates human demonstration videos, failing repeatedly without pain, gradually improving through trial and error. This process captures raw simulation footage, highlighting the beauty and terror of algorithmic persistence. The collection explores themes of ambition, fear, control, uncertainty, and humanity's relationship with accelerating technology. Released on the Ethereum blockchain via Fellowship, a gallery specializing in technology-driven contemporary art, it emphasizes blockchain-mediated distribution, audience participation in auctions, and the visibility of creative processes amid AI dominance. Each piece represents a unique moment in the avatar's journey, turning experimental failure into collectible art.