Farming Didn't End When the Tractor Arrived

Tractors did not end farming. They ended a specific version of it — the version where the binding constraint was how much ground a person and an animal could work between sunrise and dark. Farmers who picked up the machine early spent the next generation learning things the job had never required before: engines, credit, yields at a scale nobody could eyeball, markets that suddenly reached further than the county.

The craft did not vanish. It moved. Knowing your soil still mattered. Reading weather still mattered. Deciding what to plant still mattered more than anything the machine could do, because the machine had no opinion about what should be in the field.

The engine and the sailor

Same story at sea. Steam and then diesel ended the era where a crew's core skill was extracting motion from wind. It did not end seafaring — the tonnage moving across oceans went up by orders of magnitude. Navigation, seamanship, judgment in bad weather, knowing when not to sail: all still the job. The sailors who treated the engine as an insult stayed on shrinking routes. The ones who learned it ran bigger ships.

Notice what the new machine actually did in both cases. It removed the part of the work that was pure effort and left the part that was judgment. Then it raised the stakes, because now one person's decision moved a hundred times more cargo.

Software is having its tractor moment

Typing code was never the point; it was the toll you paid to get an idea into production. That toll is collapsing. What is left is the part that was always harder: deciding what should exist, judging whether what got built is actually correct, and standing behind it in front of users and stakeholders.

The pessimistic read — engineering is over — mistakes the toll for the trade. The naive read — the model does it all now — mistakes a fast draft for a finished ship. Both readings will be embarrassing in five years, for opposite reasons.

Why early matters

In every one of these transitions, the people who adopted the new machine early did not just get more done. They wrote the standard. They figured out which practices were safe, which shortcuts got people hurt, and what "good work" meant on the other side. The habits they formed became the training that everyone after them received as obvious.

That window is open in software right now, and it is wide open. Nobody has settled what review looks like when most of the diff is generated. Nobody has settled how you prove an AI-built feature is safe, or how you brief a model on a messy stakeholder request without laundering the ambiguity into confident nonsense. Those answers are being written this year, by people shipping this year.

If you are new — a first-year engineer, a PM who started building, someone rebuilding a career on purpose — this is the least disadvantaged you will ever be relative to people with fifteen years of muscle memory in the old constraint. That is not a reason to skip the fundamentals. It is a reason to acquire them fast, while the machine handles the effort.

The craft you keep

Taste, judgment, ownership. Say what you meant to build. Prove it works. Answer for it when it doesn't. That has been the job through every machine, and it is the job now.

Build. Ship. Learn.