Case Study: Prompt Partners — AI-Supported Note Formatting App

Team Members: Liam Giraldo, Tenny George, Troy Corbitt, Raymond Xu
Course: INFO 4340: App Design and Prototyping – Cornell University
Duration: 6 sprints
Role: UX Researcher / Frontend Contributor / AI Prompt Tester
Tools: Vue, JavaScript, Tailwind CSS, CodeMirror, localStorage, jsPDF, local LLM, Figma, GitHub Pages
Live Prototype: https://cornell-info4340-2026sp.github.io/prompt-partners-project/
Keywords: AI note-taking, Markdown, HTML, local LLM, prompt design, user testing, export tools, human-AI interaction

🧠 Overview

Prompt Partners is an AI-supported note formatting app that helps users turn rough notes into cleaner, structured output. The app was designed around a common student workflow: writing quick, messy notes first, then needing to organize, format, save, copy, or export them later.

Instead of using AI as a chatbot, we used it as a formatting assistant. The app lets users type notes, receive AI formatting suggestions, preview the formatted result, switch between visual and code views, save notes, and export their work in multiple formats.

The goal was to support the user’s writing process without taking control away from them.

🎯 User Task

Main task:
Help users convert rough notes into organized Markdown, HTML, or downloadable output.

Supporting tasks:

• Write or paste notes
• Apply line-level formatting
• Use AI formatting suggestions
• Preview formatted notes
• Switch between formatted preview and code preview
• Save and reopen notes
• Copy output
• Download notes as text, Markdown, HTML, or PDF

🔬 Problem Space

Many users take notes quickly, but rough notes are often hard to reuse. Formatting notes manually into Markdown, HTML, or a clean document adds extra work and interrupts the writing flow.

Our project focused on three core problems:

1. Users need help structuring messy notes.
2. AI suggestions need to be helpful without feeling too aggressive.
3. Notes need to be reusable outside the app through saving, copying, and downloading.

🧩 Prototype App

Prompt Partners includes a writing area, AI formatting support, preview window, saved note system, and export tools.

FeaturePurposeText editorLets users write and edit notesAI suggestionsRecommends formatting for note linesSelection toolbarSupports bold, italic, underline, and linksPreview toggleSwitches between formatted and code viewsNote storageSaves and reloads notes using localStorageCopy/downloadLets users reuse notes outside the appPDF exportAllows formatted notes to be downloaded as PDFs

The backlog included core interface work such as designing the text input, preview interface, code preview, Markdown/HTML toggle, note storage, and AI suggestions.

🤖 LLM Support

The LLM was used to suggest formatting, not to write the full note. Early sprints tested whether AI could identify headings, paragraphs, list items, and important words from user input.

Over time, we refined the prompts because the AI sometimes over-formatted content or became too constrained. By the final sprint, the team added built-in examples and structured AI test cases to make suggestions more consistent.

A key lesson was that note-specific constraints were too limiting. The team found that giving the AI strict rules by note type made outputs feel repetitive, so those constraints were simplified or removed.

🔁 Data Flow

The app follows a simple flow:

1. User writes notes in the editor.
2. The app stores the note as structured line data.
3. AI suggests formatting for lines or selected text.
4. The user can accept, reject, or manually adjust formatting.
5. The preview updates based on the current note state.
6. The user saves, copies, or downloads the output.

Vue components communicate through props and emitted events. For example, the text input updates the main note state, the preview receives that state, and the storage/export components use the same data to save or download notes.

🧪 Sprint Process

Sprint 1: Project Direction

Defined the app concept and focused on AI-supported note formatting.

Sprint 2: Technical Prototype

Built the first functional prototype with a text area, AI-formatted preview, per-word suggestions, per-line formatting suggestions, and basic note storage.

Sprint 3: Interaction Refinement

Explored better editing interactions, formatting controls, and preview behavior.

Sprint 4: Core Workflow

Improved the full workflow of writing, formatting, previewing, saving, copying, and downloading.

Sprint 5: User Testing

Tested usability, AI suggestion clarity, and whether the app felt intuitive.

Sprint 6: Final Polish

Added PDF export, built-in AI examples, structured test cases, prompt refinements, final testing, and code cleanup.

📊 Key Findings

1. AI Help Should Be Optional

Users benefit from AI formatting, but they still need control. Accepting, rejecting, or editing suggestions matters.

2. Formatting Needs Clear Feedback

Users need to understand what changed, what view they are in, and whether their note was saved or downloaded.

3. Export Features Add Real Value

Copy, download, and PDF export made the app feel more practical because users could move their notes elsewhere.

4. AI Needed Testing Like Any Other Feature

The team created structured test cases because it was hard to judge AI accuracy informally.

🧩 Design Implications

Prompt Partners showed that AI note tools should:

• Keep users in control
• Make AI suggestions visible and reversible
• Support both visual preview and code preview
• Allow saving and exporting
• Avoid over-constraining AI with rigid prompt rules
• Test AI behavior with structured examples

🔍 Limitations

• The app relies on a local LLM, which can take time to load.
• AI suggestions may still be inconsistent.
• localStorage works for a prototype but is not the same as account-based cloud storage.
• Testing was limited to a small group of users.
• Advanced formatting and PDF layout options could be improved.

📈 Future Work

Future improvements could include:

• Better explanations for AI suggestions
• More precise accept/reject controls
• Search across saved notes
• Cloud syncing
• Stronger grammar and spell-check support
• Improved PDF formatting
• More advanced personalization based on user behavior

📌 Conclusion

Prompt Partners explored how AI can support note formatting without replacing the user’s writing process. Across six sprints, the project evolved from a simple technical prototype into a more complete app with AI suggestions, preview modes, note storage, copy/download tools, and PDF export.

The main takeaway is that AI works best as a support layer. It should help users structure and reuse their notes, while still leaving them in control of the final result.

‍

‍

Tools: Vue, JavaScript, Tailwind CSS, CodeMirror, localStorage, jsPDF, local LLM, Figma, GitHub Pages
Live Prototype: https://cornell-info4340-2026sp.github.io/prompt-partners-project/

🧠 Overview

Prompt Partners is an AI-supported note formatting app that helps users turn rough notes into cleaner, structured output. The app was designed around a common student workflow: writing quick, messy notes first, then needing to organize, format, save, copy, or export them later.

Instead of using AI as a chatbot, we used it as a formatting assistant. The app lets users type notes, receive AI formatting suggestions, preview the formatted result, switch between visual and code views, save notes, and export their work in multiple formats.

The goal was to support the user’s writing process without taking control away from them.

🎯 User Task

Main task:
Help users convert rough notes into organized Markdown, HTML, or downloadable output.

Supporting tasks:

• Write or paste notes
• Apply line-level formatting
• Use AI formatting suggestions
• Preview formatted notes
• Switch between formatted preview and code preview
• Save and reopen notes
• Copy output
• Download notes as text, Markdown, HTML, or PDF

🔬 Problem Space

Many users take notes quickly, but rough notes are often hard to reuse. Formatting notes manually into Markdown, HTML, or a clean document adds extra work and interrupts the writing flow.

Our project focused on three core problems:

1. Users need help structuring messy notes.
2. AI suggestions need to be helpful without feeling too aggressive.
3. Notes need to be reusable outside the app through saving, copying, and downloading.

🧩 Prototype App

Prompt Partners includes a writing area, AI formatting support, preview window, saved note system, and export tools.

FeaturePurposeText editorLets users write and edit notesAI suggestionsRecommends formatting for note linesSelection toolbarSupports bold, italic, underline, and linksPreview toggleSwitches between formatted and code viewsNote storageSaves and reloads notes using localStorageCopy/downloadLets users reuse notes outside the appPDF exportAllows formatted notes to be downloaded as PDFs

The backlog included core interface work such as designing the text input, preview interface, code preview, Markdown/HTML toggle, note storage, and AI suggestions.

🤖 LLM Support

The LLM was used to suggest formatting, not to write the full note. Early sprints tested whether AI could identify headings, paragraphs, list items, and important words from user input.

Over time, we refined the prompts because the AI sometimes over-formatted content or became too constrained. By the final sprint, the team added built-in examples and structured AI test cases to make suggestions more consistent.

A key lesson was that note-specific constraints were too limiting. The team found that giving the AI strict rules by note type made outputs feel repetitive, so those constraints were simplified or removed.

🔁 Data Flow

The app follows a simple flow:

1. User writes notes in the editor.
2. The app stores the note as structured line data.
3. AI suggests formatting for lines or selected text.
4. The user can accept, reject, or manually adjust formatting.
5. The preview updates based on the current note state.
6. The user saves, copies, or downloads the output.

Vue components communicate through props and emitted events. For example, the text input updates the main note state, the preview receives that state, and the storage/export components use the same data to save or download notes.

🧪 Sprint Process

Sprint 1: Project Direction

Defined the app concept and focused on AI-supported note formatting.

Sprint 2: Technical Prototype

Built the first functional prototype with a text area, AI-formatted preview, per-word suggestions, per-line formatting suggestions, and basic note storage.

Sprint 3: Interaction Refinement

Explored better editing interactions, formatting controls, and preview behavior.

Sprint 4: Core Workflow

Improved the full workflow of writing, formatting, previewing, saving, copying, and downloading.

Sprint 5: User Testing

Tested usability, AI suggestion clarity, and whether the app felt intuitive.

Sprint 6: Final Polish

Added PDF export, built-in AI examples, structured test cases, prompt refinements, final testing, and code cleanup.

📊 Key Findings

1. AI Help Should Be Optional

Users benefit from AI formatting, but they still need control. Accepting, rejecting, or editing suggestions matters.

2. Formatting Needs Clear Feedback

Users need to understand what changed, what view they are in, and whether their note was saved or downloaded.

3. Export Features Add Real Value

Copy, download, and PDF export made the app feel more practical because users could move their notes elsewhere.

4. AI Needed Testing Like Any Other Feature

The team created structured test cases because it was hard to judge AI accuracy informally.

🧩 Design Implications

Prompt Partners showed that AI note tools should:

• Keep users in control
• Make AI suggestions visible and reversible
• Support both visual preview and code preview
• Allow saving and exporting
• Avoid over-constraining AI with rigid prompt rules
• Test AI behavior with structured examples

🔍 Limitations

• The app relies on a local LLM, which can take time to load.
• AI suggestions may still be inconsistent.
• localStorage works for a prototype but is not the same as account-based cloud storage.
• Testing was limited to a small group of users.
• Advanced formatting and PDF layout options could be improved.

📈 Future Work

Future improvements could include:

• Better explanations for AI suggestions
• More precise accept/reject controls
• Search across saved notes
• Cloud syncing
• Stronger grammar and spell-check support
• Improved PDF formatting
• More advanced personalization based on user behavior

📌 Conclusion

Prompt Partners explored how AI can support note formatting without replacing the user’s writing process. Across six sprints, the project evolved from a simple technical prototype into a more complete app with AI suggestions, preview modes, note storage, copy/download tools, and PDF export.

The main takeaway is that AI works best as a support layer. It should help users structure and reuse their notes, while still leaving them in control of the final result.

‍

‍Keywords: AI note-taking, Markdown, HTML, local LLM, prompt design, user testing, export tools, human-AI interaction
Live Prototype: https://cornell-info4340-2026sp.github.io/prompt-partners-project/

🧠 Overview

Prompt Partners is an AI-supported note formatting app that helps users turn rough notes into cleaner, structured output. The app was designed around a common student workflow: writing quick, messy notes first, then needing to organize, format, save, copy, or export them later.

Instead of using AI as a chatbot, we used it as a formatting assistant. The app lets users type notes, receive AI formatting suggestions, preview the formatted result, switch between visual and code views, save notes, and export their work in multiple formats.

The goal was to support the user’s writing process without taking control away from them.

🎯 User Task

Main task:
Help users convert rough notes into organized Markdown, HTML, or downloadable output.

Supporting tasks:

• Write or paste notes
• Apply line-level formatting
• Use AI formatting suggestions
• Preview formatted notes
• Switch between formatted preview and code preview
• Save and reopen notes
• Copy output
• Download notes as text, Markdown, HTML, or PDF

🔬 Problem Space

Many users take notes quickly, but rough notes are often hard to reuse. Formatting notes manually into Markdown, HTML, or a clean document adds extra work and interrupts the writing flow.

Our project focused on three core problems:

1. Users need help structuring messy notes.
2. AI suggestions need to be helpful without feeling too aggressive.
3. Notes need to be reusable outside the app through saving, copying, and downloading.

🧩 Prototype App

Prompt Partners includes a writing area, AI formatting support, preview window, saved note system, and export tools.

FeaturePurposeText editorLets users write and edit notesAI suggestionsRecommends formatting for note linesSelection toolbarSupports bold, italic, underline, and linksPreview toggleSwitches between formatted and code viewsNote storageSaves and reloads notes using localStorageCopy/downloadLets users reuse notes outside the appPDF exportAllows formatted notes to be downloaded as PDFs

The backlog included core interface work such as designing the text input, preview interface, code preview, Markdown/HTML toggle, note storage, and AI suggestions.

🤖 LLM Support

The LLM was used to suggest formatting, not to write the full note. Early sprints tested whether AI could identify headings, paragraphs, list items, and important words from user input.

Over time, we refined the prompts because the AI sometimes over-formatted content or became too constrained. By the final sprint, the team added built-in examples and structured AI test cases to make suggestions more consistent.

A key lesson was that note-specific constraints were too limiting. The team found that giving the AI strict rules by note type made outputs feel repetitive, so those constraints were simplified or removed.

🔁 Data Flow

The app follows a simple flow:

1. User writes notes in the editor.
2. The app stores the note as structured line data.
3. AI suggests formatting for lines or selected text.
4. The user can accept, reject, or manually adjust formatting.
5. The preview updates based on the current note state.
6. The user saves, copies, or downloads the output.

Vue components communicate through props and emitted events. For example, the text input updates the main note state, the preview receives that state, and the storage/export components use the same data to save or download notes.

🧪 Sprint Process

Sprint 1: Project Direction

Defined the app concept and focused on AI-supported note formatting.

Sprint 2: Technical Prototype

Built the first functional prototype with a text area, AI-formatted preview, per-word suggestions, per-line formatting suggestions, and basic note storage.

Sprint 3: Interaction Refinement

Explored better editing interactions, formatting controls, and preview behavior.

Sprint 4: Core Workflow

Improved the full workflow of writing, formatting, previewing, saving, copying, and downloading.

Sprint 5: User Testing

Tested usability, AI suggestion clarity, and whether the app felt intuitive.

Sprint 6: Final Polish

Added PDF export, built-in AI examples, structured test cases, prompt refinements, final testing, and code cleanup.

📊 Key Findings

1. AI Help Should Be Optional

Users benefit from AI formatting, but they still need control. Accepting, rejecting, or editing suggestions matters.

2. Formatting Needs Clear Feedback

Users need to understand what changed, what view they are in, and whether their note was saved or downloaded.

3. Export Features Add Real Value

Copy, download, and PDF export made the app feel more practical because users could move their notes elsewhere.

4. AI Needed Testing Like Any Other Feature

The team created structured test cases because it was hard to judge AI accuracy informally.

🧩 Design Implications

Prompt Partners showed that AI note tools should:

• Keep users in control
• Make AI suggestions visible and reversible
• Support both visual preview and code preview
• Allow saving and exporting
• Avoid over-constraining AI with rigid prompt rules
• Test AI behavior with structured examples

🔍 Limitations

• The app relies on a local LLM, which can take time to load.
• AI suggestions may still be inconsistent.
• localStorage works for a prototype but is not the same as account-based cloud storage.
• Testing was limited to a small group of users.
• Advanced formatting and PDF layout options could be improved.

📈 Future Work

Future improvements could include:

• Better explanations for AI suggestions
• More precise accept/reject controls
• Search across saved notes
• Cloud syncing
• Stronger grammar and spell-check support
• Improved PDF formatting
• More advanced personalization based on user behavior

📌 Conclusion

Prompt Partners explored how AI can support note formatting without replacing the user’s writing process. Across six sprints, the project evolved from a simple technical prototype into a more complete app with AI suggestions, preview modes, note storage, copy/download tools, and PDF export.

The main takeaway is that AI works best as a support layer. It should help users structure and reuse their notes, while still leaving them in control of the final result.

‍