Medellín’s city government wanted to raise its efficacy and legitimacy, especially in neighborhoods with weak state presence and competing armed actors. The city identified 80 such neighborhoods. In half, they intensified non-police street presence tenfold for two years, attempting to improve social services and dispute resolution. Unexpectedly, despite increased attention from street staff, residents lowered their opinion of the state on average. We trace these adverse effects to communities where state presence was initially weakest. Staff in these neighborhoods worked to improve services, but the central administration struggled to deliver on these promises. Where state presence was already established, however, the intervention raised opinions of the government as expected. We hypothesize that it is costlier for states to improve services where it is weak—an incentive for bureaucrats and elected leaders to concentrate statebuilding efforts in established areas, widening inequality in public service access and local variance in state legitimacy.