Working memory is small, fixed, and the thing every screen is actually competing for. Duolingo treats it as a budget with a ledger — effort that is inherent to the task, effort wasted on presentation, and effort spent on real processing — and designs each unit against that ledger rather than against a content plan.
This skill is about one screen or one sitting. The shape of the whole path — what order things come in, when supports get removed — belongs to the progression-design skill; see [[../duo-progression-design/references/scaffold-and-fade-on-a-schedule|scaffold and fade on a schedule]]. If "my onboarding is bad" is really "the first session asks too much at once," you are in the right place.
Start here
- [[references/three-kinds-of-effort-cap-cut-and-add]] — sort effort into inherent, wasted, and productive; cap the first, cut the second, add the third on purpose.
- [[references/salience-is-a-budget-and-it-front-loads]] — attention amplifies one signal and suppresses the rest, and it spends itself early in any sequence.
Sizing a unit
- [[references/cap-novelty-then-force-immediate-use]] — a hard, tool-enforced ceiling on new material per unit, and zero gap between introducing it and using it.
- [[references/hold-everything-but-the-target-constant]] — a unit that teaches one thing must freeze every other variable, or the signal has competition you created.
- [[references/isolate-the-new-element-at-the-front]] — meet the new element alone, at the start, so the later mixed task tests integration rather than recognition.
Spending the budget on the right thing
- [[references/familiar-content-is-a-load-reduction-technique]] — running a new skill over content the user knows cold redirects the whole budget onto the target.
- [[references/a-redundant-cue-suppresses-what-it-duplicates]] — when two signals carry the same meaning the cheaper one wins, so every scaffold needs a scheduled removal.
- [[references/pre-load-before-the-moment-not-after]] — five items at the threshold beat a hundred-item reference consulted after the user is already lost.
Structural moves
- [[references/give-a-hard-sub-skill-its-own-surface]] — split a competing prerequisite onto its own surface so the main flow can assume it is in progress, not blocked on.
- [[references/untailored-real-world-input-overwhelms]] — raw material is not calibrated to this user, so it produces wrong models, not just overwhelm; calibrated content is the bridge.
How to use this skill
- Name the one thing the screen or session exists to achieve.
- List everything else on it that is new, loud, or unexplained. That list is your bill.
- Cut the presentation waste, cap the inherent load by shrinking the unit, and keep only friction that changes what the user ends up able to do.
- Check the tensions. Duolingo's own posts contradict each other on bite-sized versus deep sessions, and on interesting content versus cheap-to-process content. Neither is resolved; both are live decisions for you.
Sibling skills
- [[../duo-design/SKILL]] — how a surface is composed visually, once you know what it must make salient.
- [[../duo-product/SKILL]] — intuitive-by-default and polish, the product-level version of not wasting the user's effort.
- [[../duo-experimentation/SKILL]] — how to test whether a load reduction actually changed an outcome.
Sources
All claims trace to posts on blog.duolingo.com, cited in the dated form inside each node's "What Duolingo does" section; slugs are verified against scripts/sources.json.