Anonymous assessment, introduced to higher education over the last twenty-five years to reduce attainment gaps, is a now common place. This paper suggests some ways in which anonymous assessment could be reconceptualised. We argue that there is scant empirical evidence of anonymity having worked in reducing attainment gaps in higher education. It encourages a student-teacher relationship characterised by mutual loss of trust. The rise of AI-powered language models urges a different perspective on who (and/or what) is carrying out learning activities, and how. Our provisional contentions are that we need to know 'the work, not the words' of learning processes; and that as university instructors we can afford to have an ambitious perspective on generative AI. Finally, we ask, when AI can support mutual knowledge-creating processes, how much longer it will make sense to work with notions of individual success, knowledge as a possession, and gradated outcomes based on fixed criteria.