For decades, workers have outsourced production to machines. In this talk, I examine one specific type of outsourcing: the “blackboxing” of knowledge production, which I define as the outsourcing of learning to machines. Instead of spending years in training, knowledge workers can now access prior knowledge by simply pushing a button of a software program that embeds that knowledge. Such blackboxing technologies can serve as useful shortcuts to the knowledge frontier; they free workers from having to learn the minute details of old knowledge and to focus instead on producing new ideas. Yet, these blackboxing technologies come with tradeoffs: they may create silos that lead knowledge workers into thinking within the frameworks embedded in the ready-made software programs, when they should be precisely thinking outside the box. Using the setting of structural biology, I find that while the arrival of a blackboxing technology increased the quantity of innovation, it also led to knowledge workers herding around solving similar, less impactful problems.