A fascinating chart from Anthropic shows how quickly AI coding systems are evolving. Most people focus on AI handling simple coding tasks, but that story is already old news. What stands out here is the dramatic improvement in solving open ended problems. These are the messy, ambiguous challenges that typically require creativity, judgment, and multiple steps of reasoning. Over the past several months, Claude Code’s success rate on these tasks surged from roughly 10 to 15 percent to more than 75 percent. At the same time: • Trivial tasks climbed toward 85 percent success • Routine tasks approached 90 percent success • Substantial tasks reached the mid 80 percent range This suggests AI is progressing beyond being a productivity tool and moving toward becoming a true collaborator capable of handling increasingly complex work. The long term implication is profound. Software development may gradually evolve from writing code to defining goals, constraints, and outcomes while AI h...