Programming small fault-tolerant quantum computers will involve challenges and constraints that are very different from the ones we face when designing algorithms for today’s devices. This makes it difficult to design algorithms or determine when particular applications will be viable without a deep knowledge of the machinery of quantum error correction. We propose a heuristic cost model for a 2D surface code architecture based on an approximation we call the “FLuid Allocation of Surface code Qubits”. We argue that this cost model generates reasonable estimates for the runtime and error rate of a quantum algorithm without requiring that the user explicitly compile the circuit into native surface code operations. Using the FLASQ model, we analyze the cost of competing with state-of-the-art classical methods for simulating quantum dynamics.