A research paper on GitHub Copilot and GPT 3.5’s productivity titled, ‘The Effects of Generative AI on High Skilled Work: Evidence from Three Field Experiments with Software Developers’ came out earlier this week. It showed that efficiency among developers grew by 26% while the number of code completions increased by 38%.
The paper also highlighted that entry-level engineers leverage coding assistants, like GitHub Copilot, more than senior developers. In this context, people on X discussed how if the productivity gains of Sonnet 3.5 and Cursor (which is touted as the newer, and more successful combination) are also studied, the results would be even more encouraging for the developer community.
Founded by Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger of Anysphere and MIT, Cursor started with the goal to write the world’s software. Its flagship product, Cursor’s popularity grew manifold recently, due to its features and AI scientist Andrej Karpathy’s …