If your code leaked tomorrow, would it matter? (maxfield.lol)

The article examines the diminishing effectiveness of closed-source software in protecting proprietary code, particularly in light of recent advancements in AI. A recent experiment by Epoch AI and METR demonstrated that a model could replicate 56% of 25 software programs, utilizing only the running application, documentation, and tests. This scenario highlights a shift in competitive dynamics: where once closed-source offered a significant delay before competitors could recreate a product, AI capabilities are shortening that gap dramatically. The implications suggest that while traditional protection through secrecy remains, competitors are increasingly equipped to clone features and functionalities efficiently from the user-facing product. Moreover, as software becomes increasingly hosted rather than installed, the lines between proprietary (usage restrictions) and closed (access restrictions) blur. While proprietary software protects data and user relationships, its defensive stance against cloning is weakening. The article posits that for many companies, a leak of their code may no longer be catastrophic, as the true value may lie in user data and established vendor relationships rather than the code itself. As AI and machine learning continue to evolve, firms that relied solely on code secrecy may face challenges, forcing a reevaluation of what truly constitutes competitive advantage in the software landscape.