We apply signalling theory to a cybercrime forum to explore how argot (slang and jargon) is used to signal trust in untrustworthy environments. We develop an argot detection tool, using word embeddings from forum and non-forum datasets, which are aligned using training annotations. Compared with prior work, our approach improves performance, with an increase in the F1 and accuracy scores. Using the detected argot to create per-user variables, we find a negative correlation between the use of argot and reputation votes. We explore the trajectories of groups of forum members to observe how the use of argot and user reputation in the forum varies over time. Our findings indicate forum users are using argot to overcome the cold start problem, a conundrum faced by new users to social networks with ranking systems and marketplaces with feedback systems. A significant group of long-standing users is characterised by high levels of argot in their early forum postings. This decreases once reputation metrics increase. This particular trajectory group are amongst the highest-rated long-term members.