Every change sent me back to redo the math

One small change costs a week because nothing stands alone

Because in a design calculation, nothing stands alone. Take the pavement design for a new road. The design traffic drives the number of standard axles the road must carry over its life. That load, set against the subgrade's bearing capacity, drives the thickness of every layer, from the asphalt courses down to the subbase. The layers drive the drainage, the earthworks quantities, the cross-sections, the tables, the figures and the text of the report. Change the design traffic, and there is no such thing as updating one number.

And the design traffic will change. A road moving through design is supposed to move: the alignment is adjusted, a fresh count revises the forecast, the client changes the design life the pavement must last. Each revision is somebody doing their job well. Downstream, each one means the same thing: back to the start of the chain, by hand. Redoing the math is only the visible part. New inputs go in, the standard models run again, the result tables change, the text that interpreted the old tables gets rewritten, the figures get renumbered, and a colleague has to check it all a second time. The chain does not care whether the change was a slightly heavier forecast or an entirely new alignment. It runs at the same speed either way: human speed.

What the loop actually costs

The cost is easy to miss because it hides in the overrun. Designs like these routinely run over their hour budget, and not because the engineering is hard: the models are standard, the method documented. What runs long is propagation, the same calculations redone by hand after every revision.

And each redo is checked again by colleagues with deliveries of their own, until the checking stops keeping up. The errors that survive peer review are not carelessness; they are the residue of a loop running faster than humans can verify. The same pattern, in a neighbouring deliverable, is in where a stormwater study's week goes.

A change costs more the later it lands

The construction industry has known for decades that the timing of a change decides its price. Patrick MacLeamy, the architect behind the effort curve that bears his name, put the rule in one sentence:

Once a project is under construction, changes become exponentially more expensive than they are in the design stage.

The curve is usually read as an argument for deciding early. But it says something about the design stage too. Design is where changes are supposed to be cheap, and even there, on a fixed fee, each one costs days of manual recalculation that nobody pays for. So the incentives quietly invert. A better idea arriving late is not good news; it is a threat to the budget. Alternatives go unexamined because evaluating them costs a week. The design freezes not when it is right, but when nobody can afford to touch it again.

There is also a cost the curve does not show. Ask engineers what wears them down and few say hard problems; hard problems are the reason they chose the profession. What wears them down is watching finished work turn worthless. A week of careful calculation, invalidated by a short email on a Tuesday. Then the same week over again. The loop is expensive, and it is demoralizing in a way difficulty never is.

When recalculation is nearly free, a change becomes a question

Re-running the calculation chain can become close to a button. Yesper, working as an AI colleague, carries the propagation: new inputs in, models re-run, tables and text updated, the whole chain consistent again in minutes instead of a week. The engineer's job does not shrink to pressing it. Every result still gets reviewed, and the engineer still signs; nothing about the scrutiny gets lighter. What disappears is the manual carrying of one changed number through the hours of downstream calculation and writing.

The shift it points to is larger than saved hours. When a revision costs a week, every change is a threat: to the budget, to the timeline, to the evenings of whoever carries the math. When a revision costs minutes, a change becomes a question you can afford to ask. What if the design traffic rises after all? What if the subgrade proves weaker? Run the chain and look. On a fixed fee, that is the difference between defending one design and comparing several: one extra design iteration per project, at the same fee, with the same scrutiny on every number. A change stops being a threat. It becomes exploration, and exploration is where better designs come from.