Some problems are easier if you reverse the edges and process in reverse topological order. Instead of asking "what can I reach from here?", you ask "what can reach me?"
For example, suppose you want the number of nodes that can reach each node. Reverse all edges and run the standard forward DP. Another case: if you need the latest possible start time instead of the earliest finish time, reverse the DAG and compute from the end.
The pattern is: reverse edges, compute a new topological order on the reversed graph, and run your DP forward on that reversed order. This runs in the same time and space as the forward version.