When evaluating a [scientific] claim: Extraordinary claims require extraordinary evidence. This is basic Bayesianism, without the calculations. For example, if a preprint is called Solution to the cosmological constant problem, one glance at the abstract is sufficient to dismiss it with high confidence.

When designing software (or anything else): The cheapest, fastest, and most reliable components of a computer system are those that aren't there. Specifically, this means ruthlessly pruning my original design, until only the necessary functional parts remain. The Ideal Final Result approach is quite useful there.

When designing anything new: Build one to throw away, or at least be prepared to (sunk cost avoidance). This one is rather controversial, with people arguing that Agile Development/RAD make this unnecessary. However, if you notice how often a major OSS project is redesigned way late in the game because the original architectural and design decisions no longer apply, you might as well plan for it in advance.