Turn a vague intention into one small step you can start today.

A guided learning prototype that understands the person before it arranges the content.

JULY 2026PRODUCT / LEARNINGAWS HACKATHON — PRESSPLAY ACADEMY CONCEPTWORKING PROTOTYPE

THE 45-SECOND READ

PROBLEM

Recommendation begins before the user can name what they need.

DECISION

Ask three questions, then let the user correct the system's understanding.

PRODUCT

A seven-lesson journey held inside one conversation workspace.

STATUS

Working prototype. Attribution and live outcome verification are still pending.

THE PRODUCT THESIS

The content is not the first answer. Understanding is.

AI Builder delays the catalogue match until it can describe the user's present state, desired direction, and smallest credible next move.

THE INTERVIEW / 01

Three questions create a shared understanding.

The prototype asks where the person is now, what they are feeling, and where they hope the path will lead. It then states its interpretation so the person can confirm, amend, or reject it before a journey is built.

Three-questions capture — supplied presentation

THE JOURNEY / 02

From an unclear thought to a seven-lesson path.

01Choose a direction

02Answer three questions

03Confirm the interpretation

04Review the journey proposal

05Begin a seven-lesson syllabus

THE WORKSPACE / 03

One surface carries the conversation, the journey, and the syllabus.

Learning-workspace capture — supplied presentation

The left rail preserves multiple journeys. The centre moves from clarification into lesson content. The right rail keeps the seven-lesson structure, sources, and progress visible without turning the conversation into a one-off answer.

EVIDENCE LEDGER / 05

What the supplied artifacts show.

— A working three-question diagnostic and editable understanding step— Journey proposals and seven-lesson syllabus views— A three-column learning workspace with conversation history— A Next.js 16 / React 19 prototype using AI SDK and assistant-ui— An AWS deployment path using ECS Fargate, a load balancer, and Bedrock

What this page does not claim yet.

— The presentation's legacy percentages are comparison baselines, not results for this prototype— No production A/B result or live learner outcome has been verified— Exact individual contribution and team attribution still require confirmation— The linked source repository is private and requires granted access

SOURCE REGISTER

Real artifacts. Marked limits.

Product captures come from the supplied AI Builder presentation. Implementation details were checked against the supplied private GitHub repository. No generated photography is used on this case page.

OPEN PRIVATE REPOSITORY ↗ACCESS REQUIRED · MASTER SNAPSHOT REVIEWED 23 JULY 2026