Code acceleration: Rethinking modern programming practices

Preprint, 2025

Using large language models to write code is a common practice among both novice programmers and experts. Multiple prompts often produce large amounts of text and code. For novice programmers, such practices can result in tangled and overly long scripts that are incomprehensible to beginners and unintuitive for experienced developers. Is there a sweet spot for the amount of generated code, or should we rethink and redesign these user practices? Here, I explore speculative futures of programming through turn-based text generation using large language models, and I advocate for a material-centered approach to designing interactions.