The rapid rise of scientific literature on deep learning in computational mechanics illustrates its potential but also its pitfalls due to its hyped-up over-optimism. This chapter summarizes the methodological overview of the preceding chapter, including assessments. These pinpoint the most promising research directions and highlight which to avoid. In addition, clear guidelines within deep learning for simulation are provided. In particular, the well-known but under-utilized principles of good scientific practice are reiterated in the context of deep learning in computational mechanics. Following these ensures much-needed transparency and integrity, thereby accelerating future improvements in this field—creating a pathway for deep learning to contribute meaningfully to computational mechanics.