Outgrow
Build Week judge journey
One learning system, two connected loops

Turn real goals into personal paths—and expertise into living education.

Learning is scattered across courses, documents, notes, projects, and AI chats. Outgrow turns a learner’s real goal and material into one path that learns from their work, then helps creators improve future versions from consented evidence.

01Deterministic sample · resettable

Try the learner journey

Start as Maya, build a misleading but valid chart, respond to Ava, watch the path adapt, and finish fresh work without help.

  • No account or API key
  • Fresh signed learner isolation
  • Reset and resume are explicit
02Persisted Forge proof · real records

See education improve through Forge

Inspect expert evidence, reviewed Topic v1, a completed learner activity, the evidenced “When totals mislead” improvement, and a new learner’s Topic v2 path.

  • Completed result opens immediately
  • Topic v1 stays intact
  • Every learner link is persisted
See education improve through Forge Guided replay

Completed inspection and guided replay use separate signed scopes, so resetting one never erases the other.

Why Outgrow is different

The path changes from evidence, not from another chat.

  1. Real work, not only answers.Studios and activities reveal what a learner can actually do.
  2. Help that deliberately fades.Ava asks first, offers one bounded hint, then steps out of fresh work.
  3. Visible, evidence-backed change.The learner sees what moved, why it moved, and which work caused it.
  4. Expertise that can improve safely.Creators review a new version while every existing learner keeps the version they started.

Outgrow turns any goal into a personal path—and turns human expertise into education that improves with every learner.

Secondary evidence

How this works

Product proof first. Technical evidence is available here when you want it.

Audited live release evidence

GPT-5.6 evidence

6 / 6 · no fallback
Goal and source → personal pathgpt-5.6-sol
Operation
path.plan.v1
Schema
learning-path.v1
Validation
Passed
Fallback
Not used
Source references
source-command-palette-accessibility
Learner-visible effect
Created a four-step source-grounded path ending in fresh independent work.
Source-grounded activitygpt-5.6-sol
Operation
activity.generate.v1
Schema
activity.v1
Validation
Passed
Fallback
Not used
Source references
source-command-palette-accessibility
Learner-visible effect
Created “Map the command palette focus lifecycle” inside the Learning Room.
Ava coaching from learner evidencegpt-5.6-sol
Operation
event.interpret.v1
Schema
ava-decision.v1
Validation
Passed
Fallback
Not used
Source references
source-command-palette-accessibility
Learner-visible effect
Asked a diagnostic question before offering a bounded hint.
Evidence-backed path changegpt-5.6-sol
Operation
path.rewrite.v1
Schema
path-rewrite.v1
Validation
Passed
Fallback
Not used
Source references
source-command-palette-accessibility
Learner-visible effect
Changed the path with a learner-readable reason and retained the Solo Challenge.
Expert evidence → Learning Blueprintgpt-5.6-sol
Operation
forge-blueprint.v1
Schema
learning-blueprint.v1
Validation
Passed
Fallback
Not used
Source references
10000000-0000-4000-8000-000000000001 · 20000000-0000-4000-8000-000000000001 · 20000000-0000-4000-8000-000000000002 · 20000000-0000-4000-8000-000000000003 · 30000000-0000-4000-8000-000000000001 · 30000000-0000-4000-8000-000000000002 · 40000000-0000-4000-8000-000000000001 · 40000000-0000-4000-8000-000000000002 · 40000000-0000-4000-8000-000000000003 · 40000000-0000-4000-8000-000000000004 · 40000000-0000-4000-8000-000000000005 · 50000000-0000-4000-8000-000000000001 · 50000000-0000-4000-8000-000000000002 · 50000000-0000-4000-8000-000000000003 · 50000000-0000-4000-8000-000000000004
Learner-visible effect
Extracted five skills and five executable lessons from demonstrated work.
Learner pattern → Topic v2 proposalgpt-5.6-sol
Operation
forge.curriculum-revision.v2
Schema
curriculum-revision.v1
Validation
Passed
Fallback
Not used
Source references
20000000-0000-4000-8000-000000000002 · 20000000-0000-4000-8000-000000000003
Learner-visible effect
Proposed “When totals mislead” while preserving Topic v1.
Concrete Build Week contribution

Codex contribution

  1. Used an isolated primary worktree and parallel specialists for product, Forge, tests, and submission evidence.
  2. Converted a local demonstration into PostgreSQL-backed arbitrary learner workspaces.
  3. Built the shared learning-intelligence and Product Experience adapter boundaries.
  4. Implemented disposable Coding Studio workspaces and the complete Forge v1→v2 lineage.
  5. Found and repaired ownership, restoration, migration, and no-fallback evaluation defects.
  6. Generated cross-browser, database, accessibility, backup, and isolated-restore checks.
  7. Built, tagged, deployed, and verified exact release images through self-hosted CI.
Release-gate coverage

Verification evidence

4 proof layers
Unit + database tests
240
Deterministic evaluations
15
Live no-fallback operations
6
Applicable browser tests
83
Exact release identity

Release evidence

v0.3.1
Release
v0.3.1
Commit
1e5646c504e6b439d017d6b0e7bebe64cd43d976
CI result
Passed for this exact deployment
Open CI evidence Open deployment evidence