ReasonMap

Strengthen your reasoning. Keep the decision yours.

ReasonMap is a guided reflection app for decisions, uncertainties, and questions without obvious answers. It helps a user examine what matters, evidence, assumptions, affected perspectives, and gaps—one question at a time. It reflects the user's thinking without making the decision for them.

Live demo

https://reasonmap-build-week.hildegardcat.chatgpt.site

The experience

ReasonMap leads the user through seven stages:

1. Frame the question
2. Identify what matters
3. Examine the evidence
4. Surface assumptions
5. Consider perspectives
6. Check for reasoning gaps
7. Reflect and summarize

The user can change the kind of support at any point:

• Guide Me offers balanced prompts.
• Challenge Me gently tests the reasoning.
• Let Me Think pauses without losing the user's place.
• Help Me Organize My Thoughts arranges what the user has shared without adding a conclusion.

The live reasoning map contains only ideas the user has confirmed. Sections remain compact until there is something useful to show, reducing overload and defensiveness.

Privacy and scope

• No account is required for the Build Week demo.
• ReasonMap does not save responses to an application database.
• The current prototype keeps the session in browser memory and clears it when the session ends or the page is refreshed.
• Users are reminded not to enter unnecessary identifying details.
• ReasonMap is a reflection tool, not professional, medical, legal, or financial advice.

Built with

• Codex with GPT-5.6
• TypeScript
• React 19
• Next.js-compatible components
• Vinext and Vite
• Tailwind/PostCSS tooling
• Cloudflare-based OpenAI Sites hosting

Run locally

Prerequisites

• Node.js 22.13 or newer
• npm
• Linux development environment (the included lifecycle scripts use GNU flock and timeout)

Setup

npm ci
npm run dev

Open the local address printed by the development server.

Production validation

npm test

This builds the application, validates the deployable artifact, and runs the rendered-page test.

Suggested judging path

1. Select Start a reasoning session.
2. Enter a decision such as: Should our program continue offering a low-enrollment course?
3. Try Guide Me, then switch to Challenge Me.
4. Add responses during the stages and watch the reasoning map update.
5. Use Let Me Think, then choose I'm Ready, Give Me a Prompt, or Move On.
6. Continue to the final reflection and expand the captured sections.
7. Choose Clear session or refresh the page to remove the current session.

How Codex and GPT-5.6 were used

Codex was a development collaborator throughout Build Week, not a one-time code generator. It helped translate a detailed educational design into a working product, implement the responsive interface, build and revise the staged session flow, test interactions, and diagnose issues found during hands-on review.

GPT-5.6 was especially useful in parts of the build that required maintaining subtle product constraints across many components:

• preserving user agency instead of producing a recommendation;
• asking one useful question at a time;
• supporting user-controlled guidance, challenge, silence, and organization modes;
• separating internal reasoning structure from what should be revealed in the interface;
• making the reasoning map gradual and nonjudgmental;
• removing repetitive or pedantic prompts after testing;
• keeping the final reflection faithful to the user's words.

Key product decisions remained human-led. The creator chose the name, purpose, method, modes, prompt principles, privacy posture, and the rule that people learn more deeply when they make discoveries for themselves. Codex accelerated the conversion of those decisions into an integrated interface and helped iterate quickly when the experience did not yet feel right.

Current prototype boundary

The Build Week version demonstrates the complete interface and interaction model with a deterministic prompt engine. A planned next step is to connect the established reasoning-state design to live GPT-5.6 responses while preserving the same safety, privacy, pacing, and user-agency constraints.

Project structure

• app/page.tsx — staged reasoning flow, modes, map, and reflection
• app/globals.css — responsive visual system
• app/layout.tsx — application shell and metadata
• tests/ — rendered application checks
• .openai/hosting.json — OpenAI Sites deployment configuration

License

Copyright © 2026 Kate Thomas. All rights reserved unless a separate license is added to the repository.