I would wager that over the last several decades BLAST (basic local alignment search tool) has done more to advance our knowledge of biology than any other algorithm.
https://bioinformaticsreview.com/20210503/how-blast-works-co...
"The heart of many well-known programs is a dynamic programming algorithm, or a fast approximation of one, including sequence database search programs like BLAST..."
https://html.duckduckgo.com/html?q=%22%20heuristic%20dynamic...
I'm clearly not going to understand your motivation so I think I'll stop here.
(edit: There is absolutely nothing stopping heuristics and DP being combined; in fact they have to be in eg. database optimisation. A pure DP solution to query optimisation will be optimal, but will take unbounded (lthough finite) time, which is unacceptable. DP and heuristics are combined to both guide the DP search and bound it after a strict time (usually a couple of seconds CPU) by when it is hopefully 'good enough').
But you’d have to look pretty close to see that’s what I did because unlike him I’m not obsessed with people thinking I’m clever. I’d rather be seen as wise.
(I started fixing it because we had found a couple mystery bugs caused by his solution not being reentrant. By the time I had full code branch coverage, it was actually six separate bugs I fixed)
There are two key attributes that a problem must have in order for dynamic programming to be applicable: optimal substructure and overlapping sub-problems. ... Optimal substructure means that the solution to a given optimization problem can be obtained by the combination of optimal solutions to its sub-problems.
See the words optimal and optimization?
This is in the site guidelines: https://news.ycombinator.com/newsguidelines.html.
If you, by optimization mean that there's a function value you are maximizing or minimizing, than that's not true either, since Subset Sum and Hamiltonian path are canonical decision problems for which DP is used.
Heck, you can even take the standard TSP DP algorithm and, instead of looking at all possible "exit vertices" in linear time, look at a randomly chosen constant number of candidates, thereby reducing the running time by a factor n and getting a randomized heuristic function not guaranteed to give the optimal value.
This reminds me of biologists bickering about what is or is not a gene, and endless snorefests of ontologists bickering about the semantics of a label on an edge in a graph.
It was not amusing then either.