As a response to the trends of the increasing importance of computational approaches and the accelerating pace in science, I propose in this position paper to establish the concept of "science bots" that autonomously perform programmed tasks on input data they encounter and immediately publish the results. We can let such bots participate in a reputation system together with human users, meaning that bots and humans get positive or negative feedback by other participants. Positive reputation given to these bots would also shine on their owners, motivating them to contribute to this system, while negative reputation will allow us to filter out low-quality data, which is inevitable in an open and decentralized system.