The best antidote I've found is shipping smaller increments. When you're betting on a 6-month feature, it feels like gambling. When you're shipping weekly, you get feedback fast enough to course-correct.

The dopamine hit from "will this work?" becomes "did users like this?" which is a healthier loop.