Identical agents exert hidden effort to produce random quantities of a public good, and receive a stock- and effort-contingent flow of payoffs. I characterise the unique symmetric Markov equilibrium as well as, for the two-agent binary-effort case, the most efficient equilibria in strategies that condition on public and on private histories. In contrast to the social-welfare benchmark, under general conditions,continuation payoffs in the symmetric and the most efficient equilibria may drop after an increment in the stock. I show that reducing the risk of producing negligible quantities in favour of moderate ones may reduce welfare in each equilibrium, and that the ability to conceal increments contingent on their size is incentive-compatible and welfare-enhancing. In the context of an R&D alliance in which improvements of a collective technology raise the incentive to shift resources towards private activities, concealing moderate improvements may enhance private profits and collective progress.