---
name: duo-attention-budget
description: Treat working memory as the binding constraint on every screen, session, and document — cap what competes for it, cut what wastes it, and deliberately spend what is left on the one thing you are trying to teach. Covers how much new material fits in one unit, why familiar content is a load-reduction technique, salience and primacy effects, redundant cues that suppress what they duplicate, priming right before the moment, isolating a new element, splitting out a hard sub-skill, and why raw real-world input overwhelms. Use when someone says my onboarding is overwhelming, users keep missing the important thing on this page, how many new concepts per screen, how long should a session be, why does nobody read my tooltips, or should I just drop users into the real data.
license: MIT
metadata:
  author: HKTITAN
  version: "2.0.0"
  graph: true
---

# Duolingo's Attention Budget

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

1. Name the one thing the screen or session exists to achieve.
2. List everything else on it that is new, loud, or unexplained. That list is your bill.
3. 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.
4. 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`.