Anthropic's new multiagent post (https://www.anthropic.com/research/multiagent-systems) found something I hadn't expected: "Only Sonnet 5 is able to maintain both a high merge fraction while directly collaborating and sharing code with other agents." Other frontier models (Opus 4.8, Mythos Preview) got a high merge fraction only by avoiding shared files almost entirely. Does anyone have a hypothesis for why Sonnet 5 can do both at once?
Because your students are non-technical. They don't have a good reference point understand to how hard building software is (IT IS HARD). It is a valid for them to come to the conclusion "I never knew that building software was so easy".