DSI is a computational model orginally based on Friedman’s and Wise’s “lazy” cons. It evolved into a model that is demand oriented (rather than demand driven) and speculative. Eric Jeschke’s parallel implementation of DSI introduced additional ideas of bounded speculation aiming at architectures that we today call “many-core”. My talk will review this work from the late 1970s to the mid 1990s, some of its novelties, and problems left open.