Stochastic thermodynamics provides a universal framework for analyzing
nano- and micro-sized non-equilibrium systems. Prominent examples are
single molecules, molecular machines, colloidal particles in time-dependent
laser traps and biochemical networks. Thermodynamic notions like work, heat
and entropy can be identified on the level of individual fluctuating
trajectories. They obey universal relations like the fluctuation theorem.
Thermodynamic inference as a general strategy uses consistency constraints
derived from stochastic thermodynamics to infer otherwise hidden properties
of non-equilibrium systems. As a paradigm for thermodynamic inference, the
thermodynamic uncertainty relation provides a lower bound on the entropy
production through measurements of the mean and dispersion of any current
in the system. Likewise, it provides a model-free bound on the thermodynamic
efficiency of molecular motors. Waiting-time distributions between consecutive
transitions in a discrete Markov network yield an even better estimator of
entropy production. Moreover, they reveal further information about the
topology of the underlying network. From the observation of coherent
oscillations, a universal bound on their thermodynamic cost can be deduced.