Universal Credit is a major change in the UK's social security system which will affect around eight million households by replacing six existing means-tested benefits and tax credits with a single benefit, based on income, assets and circumstances, and paid monthly. Much commentary about Universal Credit has supported the principle but raised concerns about delivery. This consensus about the principles and the focus on delivery have resulted in less attention being paid to key policy choices informing the design. This article examines two aspects of the design of Universal Credit: the aim to make Universal Credit as much ‘like work’ as possible, and the architecture of the means test. The focus here is the gap between the assumptions underlying the design of Universal Credit, on the one hand, and the research evidence about life on a low income, and in low-waged and often insecure employment, on the other. Finally, we discuss the most significant contradiction between the ‘transformational’ aim for Universal Credit, to help people achieve greater ‘independence’ from the state, and the realities of deepening and widening control of claimants’ lives.