Why estimate?

Analysis is cheap now. Software will return a bending moment to six significant figures for any model you give it, right or wrong. What it will not do is tell you that the model is wrong. That is the engineer’s job, and the tool for it is an expectation: a sense of roughly what the answer should be, held before the output arrives.

That sense used to be absorbed — from the right projects, the right supervisor, the right few years of checking other people’s numbers. falsewrk exists to deliver those reps deliberately instead of by luck.

What a good estimate is

A good estimate is fast, roughly right, and honest about its direction.

• Fast, because its value is in catching problems early — in a design meeting, on a site walk, while reviewing a calculation package. An estimate that takes twenty minutes is a calculation.
• Roughly right: within 10% is excellent, within 40% is a useful check. Being out by a factor of ten means something structural was missed — a unit, a load path, a span squared instead of cubed.
• Honest about direction: knowing that you tend to land high or low is almost as useful as being right, because you can correct for it.

Why this kind of practice works

Research on expert intuition points to two conditions for judgement to develop: a domain with real regularities to learn, and quick, unambiguous feedback on each attempt (Kahneman & Klein, 2009). Structural behaviour has the regularities. What practice usually lacks is the feedback — on a real project you might wait months to learn whether an assumption held, if you ever find out at all.

A rep closes that gap to seconds. You commit to a number, then see the true value and the method straight away. Hundreds of those, spaced well, are what build the instinct.

What falsewrk does not do

It does not replace design. The problems are deliberately simplified so the answer can be known exactly, and the methods shown are for building judgement, not for use in a design without checking them against the governing code.