Recently, OpenAI released a significant report on AI programming capabilities, highlighting the current state of AI in software development through a $1 million real-world development project. The benchmark test, named SWE-Lancer, covered 1,400 real projects from Upwork, comprehensively assessing AI performance in both direct development and project management areas. The results indicated that the best-performing AI model, Claude 3.5 Sonnet, achieved a success rate of 26.2% in coding tasks and reported performance in project management.