Somehow, these paradoxes don't work for me, because they arbitrarily assume that something is a sample. Take the ten card thing. If my hypothesis is that no card matches its position, then only turning over all cards counts as a sample. Turning over one card is no more confirmatory than spotting a black raven, the existence of which is wholly consistent with my ten-card hypothesis but is not thought to confirm it in the least.

The grue paradox is really two unrelated hypotheses. The first is that all emeralds will be green on any date before some stated date. That hypothesis, with respect to any date before said date that has already occurred, is confirmed by observation and induction. The second is that all emeralds will be blue after the stated date. That hypothesis cannot be tested until that date. Again, spotting a green emerald now would be no more confirmatory of the second hypothesis than spotting a black raven would be.

What am I missing?