Adjustment costs are a central bottleneck of the real-world economic transition essential for achieving the sizeable reduction of greenhouse gas (GHG) emissions set out by policy makers. Could these costs derail the transition process to green growth, and if so, how should policy makers take this into account? I study this issue using the model of directed technical change in Acemoglu, Aghion, Bursztyn, and Hemous (2012), AABH, augmented by a friction on the choice of scientists developing better technologies. My results show that such frictions, even minor, materially affect the outcome. In particular, the risk of reaching an environmental disaster is higher than in the baseline AABH model. Fortunately, policy can address the problem. Specifically, a higher carbon tax ensures a disaster-free transition. In this case, the re-allocation of research activity to the clean sector happens over a longer but more realistic time horizon, namely around 15 instead of 5 years. An important policy implication is that optimal policies do not act over a substantially longer time horizon but must be more aggressive today in order to be effective. In turn, this implies that what may appear as a policy failure in the short-run | a slow transition albeit aggressive policy | actually re ects the efficient policy response to existing frictions in the economy. Furthermore, the risk of getting environmental policy wrong is highly asymmetric and `robust policy' implies erring on the side of stringency.