Capability paths, parent feedback, and mastery — proposal

September 12, 2026. Planning/backlog only. Builds on shipped #180; parent feedback and AI planning remain owned by #146. No production behavior is changed by this proposal. These are proposed teaching paths, not validated developmental sequences or age-equivalent scores.

Product direction

Give each child an encouraging player card and a branching skill map. Children accumulate achievements and discover new challenges, rather than start with a perfect grade that falls. Separate scissors, pencil control, reading, spelling, math, and other capabilities. Age remains optional starting context for unknown areas. Parents can choose a focus, confirm what they have already seen outside Scribble Works, change support preferences, and correct observations.

Use a skill graph with several paths, not one overall child level. Distinguish three edge types: a reviewed learning foundation, a recommended next experience, and an execution demand that an adult or adaptation can support. A pencil requirement on a worksheet is not a prerequisite for knowing numbers. Listening comprehension must be available independently of decoding and handwriting. English and Spanish observations stay separate.

The Backyard Baseball reference informs the player card, visible strengths, and earned progress. Use original Scribble Works artwork. Domain bars summarize explicitly scoped achievements (for example, 3 skills earned in this path), never a percentile, age equivalent, intelligence score, or percentage of an entire developmental domain. Adding catalog nodes must not make earned accomplishments appear to disappear.

Public precedents and how to use them

• Khan Academy mastery levels: demonstrates per-skill progression through performance. Its system can also lower levels after errors; we propose keeping earned celebrations while updating current support needs separately.
• Brilliant learning paths: ordered courses, active lessons, practice checkpoints, and self-paced advancement. Borrow clear next steps and checkpoints. This is product inspiration, not validation of our proposed child-development graph.
• Brilliant product features: distinguishes activity XP and provides a parent progress dashboard. Keep practice activity distinct from demonstrated capability in Scribble Works.
• Brilliant K–5 practice: public grade/topic navigation and small math tasks. At review, kindergarten and grade 1 had quizzes; grades 2–5 were marked coming soon. Useful for task granularity, not a complete K–5 curriculum to import.
• Head Start Early Learning Outcomes Framework: a public framework of domains, goals, developmental progressions and observable indicators. Use to check breadth and source mappings, not to turn age ranges into eligibility gates or claim it is a game curriculum.
• Learning and Teaching with Learning Trajectories, Clements and Sarama, University of Denver: connects early math goals, progressions and activities. Strong candidate for math sequence review. Publicly accessible does not mean freely reusable; link/reference and author our own descriptions, and review permissions before reusing protected activities or text.
• Common Core counting and cardinality: distinguishes counting objects, number names, and quantities. Useful standards crosswalk; grade labels remain reference metadata.
• Colorado Department of Education competency-based learning resource: describes advancing through demonstration of mastery. The educational approach exists publicly; no single cited source supplies our complete proposed game tree.
• CDC developmental milestones: useful developmental monitoring context, not a curriculum or mastery-scoring system. CDC explicitly says these resources do not replace validated screening tools. Keep milestone monitoring separate from game achievements.

Research boundary: reviewed official public pages and the public Brilliant course entry point. Did not inspect a paid learner journey or purchase a subscription. No claim of a validated curriculum or learning-outcome improvement.

First tree: full mapping of the 34 existing observations

Arrows below are proposed teaching sequences to review, not automatic mastery gates. Brackets identify proposed additions or splits. Existing catalog IDs are retained here for a reliable migration crosswalk.

Broader future coverage: communication/listening, social interaction, emotion regulation, approaches to learning/persistence, and gross movement need independently defined tasks and suitable activities. Do not show empty branches as skills a child lacks. Our current printable catalog is only a partial educational offering.

Content audit of the shipped catalog

Source inspected: src/lib/learning/catalog.js at the #180 release, 34 objective observations / 39 game mappings. Six objectives currently have no directly targeted game: write_letters, text_en, fraction_parts, clock_hours, money_coins, letter_match. They appear as demands or observations but lack dedicated targeted practice in the mapping. Review actual printables before authoring new content; this is a metadata coverage finding, not proof that no page exercises them incidentally.

Other launch blockers for mastery: broad bundled objectives; shape tracing versus copying; number ordering versus cardinality; oral versus written responses; coin arithmetic versus identifying denominations; and many objectives with only one game. One repeated worksheet cannot establish transfer. Current FOUNDATIONS is documented as task demands, not a promotion ladder; do not repurpose it wholesale.

Closing the loop

1. Parent chooses one or two current focus skills; system suggests an achievable practice activity with a clear reason and explicit assistance options.
2. Household receives/prints a playset. Delivery, downloading and printing are not evidence that the child played.
3. A brief optional check-in asks: played/not yet; easy/just right/hard; alone/some help/lots of help; enjoyable/not enjoyable. Ask one targeted observable question, such as whether the child blended unfamiliar short words without someone supplying the sounds. Permit skipped/unsure answers and optional notes.
4. Save dated per-objective evidence, game/version, response mode, parent assistance, reporter, and corrections. Mixed outcomes stay separate. A parent cutting pieces gives no evidence of the child's cutting ability.
5. Update practice history and a transparent capability summary. Suggest a checkpoint or new focus when supported; explain why. Parents can accept, defer, explore an alternative, or correct it.
6. Feed the same reviewed profile into account recommendations, daily curation, AI candidate validation, fallback and swaps. Parent sees what changed in the next plan. Keep mastered favorites available for enjoyment and occasional review.

No email or lifecycle nudges are authorized by this planning item. Link feedback from account playset history first; any later email change remains explicit work.

Progress and promotion proposal

Two distinct records:

• Practice/XP: parent-confirmed play or actual observed in-product activity; rewards participation. Repetition can prompt a check-in or a different variant. Delivered/downloaded sets alone do not earn played-game credit. Repeated play alone never promotes mastery.
• Capability evidence: observed task performance and degree of assistance. Suggested display stages: Exploring → With help → Independently → Demonstrated across examples. Unknown remains unknown, not zero ability.

Pilot rule to test with an early-childhood educator: independent success on more than one occasion, plus a different example/context, followed by parent confirmation. This is a proposed product heuristic, not a scientifically validated universal threshold. Define a rubric for each objective; do not use a universal 3-plays or 80%-correct rule. Parents may report equivalent prior observations so children need not grind through skills they already demonstrate.

Progress bars show fulfilled named checkpoint criteria, not invented probabilities of mastery. An evidence detail might read “Independent example recorded; another occasion and a new example still to try.” If a numeric bar is used, label it as checkpoint completion. Keep confidence/provenance in the parent detail, not as a child-facing percentage.

When a checkpoint is confirmed, celebrate an earned badge and suggest adjacent ready nodes. “Unlocked” means ready to explore/recommended; it is not a blanket age gate or a requirement to fill every unrelated branch. Parents can choose supported exploration. Underlying assistance/safety rules still apply.

Retain historical achievements when evidence later changes, but update current readiness and suggest gentle practice. Do not erase badges for a difficult day or a missed streak; do not keep serving harder games based on stale mastery. Avoid sibling rankings, global intelligence scores, and streak penalties.

Implementation slices and acceptance criteria

Implementation stays in the backlog until selected. #146 remains the parent-feedback and AI integration owner; this issue owns the skill-tree specification and player-card progression design. Coordinate content with #142 and freshness/variants with #161.

• Review the above graph with an early-childhood educator and literacy/math specialists as appropriate. Record source, rationale, prerequisite vs recommendation vs execution-demand edges, rubric, accommodations, and content coverage for each pilot node. Version definitions and preserve old observations without upgrading their meaning.
• Specify append-only/correctable observation history and a derived current summary, with household authorization, timestamps, response language/mode, assistance, provenance and deletion behavior. Existing “solid” is a parent snapshot, not retroactive validated mastery.
• Deliver parent feedback under #146 with skip/unknown states and a visible next-plan response. Duplicate submission cannot create extra XP or independent evidence. Corrections are reflected across selectors.
• Integrate the AI picker under #146 through schema, deterministic validator, fallback and swaps. Model proposes/explains; reviewed rules govern capability eligibility and promotion. Consent/data minimization precede transmission of saved learning profiles. AI failure retains an appropriate deterministic plan.
• Prototype the player card and tree: current focus, earned milestones, next choices, checkpoint detail, support mode and an accessible list alternative. No age comparison or global child level. Empty/unknown branches explain missing observations or missing content honestly.
• Pilot a small number of complete paths with varied practice and checkpoint content before expanding all 34 observations. Record parent burden, correction/override rate, perceived challenge/enjoyment and transfer examples; do not claim efficacy from completion counts.
• Test mixed scissors/reading profiles, different languages, adult-completed parts, repeated identical games, missing feedback, stale/contradictory reports, already-known skills, changing graph versions, sparse inventory, and more advanced oral reasoning than reading/writing.

First recommended build slice after specification: feedback and evidence history → consistent picker behavior → checkpoint/level-up presentation. The card's progress needs a trustworthy feedback mechanism underneath it.