| Mobile Web

AI coding splits into before and after GPT-5.1 and Opus 4.5, engineer says

Coding agents are reaching a turning point as next-generation AI models move beyond plausible code to producing applications that actually work, an assessment said. Software engineer Simon Willison said newer models are starting to deliver results that operate as instructed, boosting trust in outputs. He said productivity has risen to as much as 10,000 lines a day with AI help, while bottlenecks shift from implementation to testing.