Kira Cui Work / Found
Independent product · AI experience

Found

From information to insight.

An AI thinking partner that helps people capture fragmented information, discover meaningful connections, and turn what they already know into new ideas.

RECORD 02 / 04 FOUND 2026.07 —
STATUS
Concept / experience prototype ● To verify
CONTEXT
Cornell Design Technology, specialization project
ROLE
Product strategy, UX research, interaction design, AI experience design
TEAM
Individual, with faculty feedback

SUBJECTSPersonal knowledge management · Human–AI interaction · Semantic retrieval

Storyboard panels: saving a text snippet, a link and a screenshot to Found from the right-click and share menus, with a small non-blocking confirmation.
FIG. 01 Storyboard, chapter 1 — capture without leaving the task.
01

The question

What if AI helped us think more deeply with what we already know, instead of simply giving us more answers?

02

Saving is not knowing

Screenshots, articles, links, chats and passing thoughts are scattered across different tools. Saving them often requires tags, folders and ongoing maintenance. Even after saving, we still have to remember what exists, search for it and make the connections ourselves.

Storyboard panels: a day of saving articles, screenshots and notes in many places, then failing to find them when starting a new project.
FIG. 02 Chapter 0 · The scattered mind — before Found.

03Mapping the landscape

  1. 01CollectAcademic papers, existing products and related resources, filtered for relevance.
  2. 02Analyse and compareFive lenses: knowledge and evidence, user intent, system intelligence, interaction, edge cases.
  3. 03Build the design spaceSynthesise the coding into structured, high-level dimensions.
  4. 04Use itMap existing tools, identify gaps, and inform directions for new PKM experiences.
Matrix of eight information-lifecycle stages — discover, capture, organize, process, connect, retrieve, reflect, create — showing where nine existing tools offer support, with their strengths and pain points.
FIG. 02Eight stages of the information lifecycle, and where existing tools stop.
Map of PKM tools by system initiative (passive to proactive) and user intent (remember to decide). Found's exploratory direction sits in the proactive, create-and-decide opportunity area.
FIG. 03System initiative × user intent. Found aims at the proactive, create-and-decide corner.
What the map shows
  1. 01Most tools stop at storing, organising or retrieving.The space is crowded with tools that help people save, search and manage content.
  2. 02AI tools are emerging, but mostly user-triggered.Current AI features usually need explicit input and are rarely integrated with personal context across sources.
  3. 03Few support the journey from understanding to action.Few solutions proactively connect past knowledge with current goals and real-world decisions.

The opportunity: a more seamless, intelligent and connected way to turn information into lasting knowledge and meaningful action.

  1. 01Easier capture across sources
  2. 02Automated organisation and AI processing
  3. 03Stronger connections across content
  4. 04Proactive resurfacing and personalised insights
  5. 05From knowledge to action
04

Where Found sits

Fifteen dimensions of personal knowledge tools. The grey line is the full range of options; the black bar is the area Found sets out to explore.

Possible optionsFound’s direction
Target usersGeneral publicProfessional expertsStudents and knowledge workers; also relevant to researchers.
Knowledge sourceSelf-createdExternalIntegrative: personal notes mixed with external sources.
Knowledge typeTextOtherMulti-modal beyond text — not audio or video for now.
Time horizonImmediateLong-term (years)Long-term accumulation and rediscovery of forgotten knowledge.
OrganisationManualAutomaticAI-assisted organisation with user control.
TransformationStoreApplyHigher-level support: connect, synthesise and apply.
User initiativeUser-initiatedSystem-initiatedProactive: the system brings relevant knowledge to the person.
InterfaceSearchAmbientMore natural and contextual interaction.
Context of useCapture in the momentReflectionReal, varied contexts — not just intentional note-taking.
EcosystemSingle appCross-platformIntegration across notes, browser, PDFs and more.
System intelligenceNo AIFully agenticAI beyond search: finding connections and synthesising.
PersonalisationGenericDeep personalisationAdaptive — learns from behaviour and evolves over time.
Privacy and ownershipThird-party cloudLocal-firstUser-controlled, privacy-respecting.
CollaborationPersonalCollaborativeMostly personal, with optional sharing.
ScaleHundredsTens of thousands+Designed for long-term growth and large knowledge bases.
05

How it works

Two loops feed one living knowledge system. People decide what to save and what to ask; AI is planned to organise in the background — and every suggestion can be accepted, refined or rejected.

LOOP 1 · KNOWLEDGE ACCUMULATIONCAPTURE · UNDERSTAND · ORGANIZE · LEARN
YOU
  1. Encounter useful information
  2. Save to Found, instantly
  3. Original is always kept
  4. Organised knowledge enters the system
AIBackground processing (automatic)Extract key content (OCR)SummariseSuggest tags and categoriesIdentify related concepts

Optional: review AI suggestions — accept, refine or reject. Never required for content to be saved.

LOOP 2 · RETRIEVAL & CREATIONRETRIEVE · EXPLORE · CREATE · ENRICH
YOU
  1. Have a question or topic
  2. Ask Found in natural language
  3. Explore, compare, develop
  4. Save the new insight
AIFind relevant notes and sources, then surface relationshipsSimilarityConnectionContradictionCreative associationEvolutionRelated contexts

New insights become part of the person's knowledge — and fuel the next question.

OUTCOME · LIVING KNOWLEDGE SYSTEMProgressive personalisation — learns from feedback and interactions.

● Target experience — implementation status and formal user testing remain to be confirmed.

06The experience, in four chapters

Storyboard chapter 1: saving text, links and screenshots to Found from any device without breaking focus.
CH. 1Capture the moment

● Storyboard of the proposed experience — not screenshots of a working product.

07Design principles

(06)
  1. 01

    Capture without interruption

    Keep saving lightweight; avoid requiring tags or immediate confirmation.

  2. 02

    Preserve the original source

    Keep source material separate from AI interpretation, so it remains accessible when an interpretation is wrong.

  3. 03

    AI organizes in the background

    Plan extraction, summaries and indexing as asynchronous work.

  4. 04

    Connections are offered, not imposed

    Show the basis for suggested links and let people accept, correct or ignore them.

  5. 05

    Retrieval should lead to thinking

    Use traceable sources to support comparison, reflection and further questions.

  6. 06

    Knowledge should stay alive

    Allow insights from a conversation to return to the personal knowledge system.

08

What I chose not to build

No. Direction Why / why not Decision
01 Manual-first organization Tags and folders at capture time add cognitive burden. Rejected
02 Confirm every AI result immediately Immediate confirmation after each save could interrupt the original task. Rejected
03 Storage-only knowledge base Archiving and search alone do not express the intended value of rediscovery. Rejected
04 AI connections without user control Connections need explanations and a way to correct them. Rejected
05 Feature-heavy first version Complex editing, tasks, collaboration, graph views and daily pushes are deferred while the core experience is explored. Rejected

09Where it stands

  1. RESEARCHLiterature and product review; nine tools compared across the information lifecycleDone
  2. DEFINEPKM design space — fifteen dimensions, with Found's directionDone
  3. SYSTEMInteraction ecosystem: two loops and a living knowledge systemDone
  4. STORYStoryboard in five chapters, from scattered to growing knowledgeDone
  5. PROTOTYPE[Working prototype — scope and status]To verify
  6. TEST[User testing — status and method]To verify
10

Still unproven

Found is a concept in progress. These assumptions still need to be tested.

  1. Open 01 Will people keep saving fragments into Found over the long term?
  2. Open 02 Does low-friction capture increase how often saved material is reused?
  3. Open 03 Are AI-suggested connections useful beyond surface-level similarity?
  4. Open 04 Can people understand, question and trust the suggested connections?
  5. Open 05 Does rediscovery help people develop more useful ideas or creative work?
11

What I’m learning

Lowering the friction of capture is only the first step. The harder question is how AI and people can understand, correct and develop knowledge together while people retain control of their thinking.