Thoughts
I'm considering creating a meta-framework for software design. Programmers are as likely as anyone to fall into sunk-cost fallacies. And
creating a working solution to hard research problems (self-driving, etc.) is itself a optimization/search problem of the type that humans are bad at solving.
Companies, especially "agile" companies focused on rapid iteration, tend to do something like hill-climbing. Constantly moving uphill keeps your software improving, but only up until you've hit a local maximum.
Maybe modern minimalist will include a recommendation for simulated annealing or something.