Augmenting a person beats automating the task

foundations

Automation asks: how do we remove the person? Augmentation asks: how do we make the person better? The difference sounds small until something unexpected happens. An automated pipeline handles every case it was designed for and quietly fails on the one it wasn't — its quality is capped at the system's blind spots, and nobody is watching the gap. A person kept in the loop catches the case the design never anticipated.

Douglas Engelbart made this the whole point of his work on "augmenting human intellect" in the 1960s: the goal was not to replace people but to extend what they could understand and do. That framing still holds. Hand the machine the volume — the repetition, the first draft, the search across a thousand files. Keep for yourself the part that needs a human: what counts as good, what's worth ignoring, when to stop.

This is why AI is one tool for thinking, not the thinker rather than a substitute for one. The tool earns its keep by lifting your ceiling, not by deciding in your place — so Delegate the task, keep the judgment.

See how it helps.

Next: practice the handoff by learning to delegate the task but keep the judgment.

This is an evergreen note — atomic, claim-titled, and densely linked — a practice from Andy Matuschak, re-implemented in our own words.

← All notes · Learning Hub