Significant Figures and Honest Precision

The number of digits we print matches the precision the analysis supports; extra digits are a form of overclaiming.

Digits are a claim about precision

Every digit printed asserts that the digit is meaningful. Printing more digits than the uncertainty justifies is a small dishonesty that compounds, because it invites readers to treat a rough estimate as a measurement. We choose significant figures to reflect the error budget, not to look authoritative.

Why some numbers carry more digits

Q_sci is quoted as 3.424 and helium-3 output as about 1.97 kg/yr because those are outputs of a settled calculation with correspondingly small model spread. A quantity dominated by an unvalidated correlation gets fewer digits and an explicit range, because more digits would be fiction. The rule is consistency between the printed precision and the propagated uncertainty.

Rounding is done once, at the end

We carry full precision through intermediate steps and round only when reporting, so accumulated rounding does not distort a result. Rounding early and repeatedly can shift a final figure by more than its true uncertainty -- a self-inflicted error that a disciplined pipeline avoids.

Ranges beat false central values

When the honest state of knowledge is a range, we report the range. The burner availability figure of roughly 0.86 to 0.995 is written as a band on purpose: collapsing it to a single number would erase the very uncertainty that matters, which is that this is 30-100x short of hyperscale Tier III at 0.99982.

Precision discipline is quiet but load-bearing. It is how the reader can trust that a number means exactly as much as it appears to.