Normalized categories
Stable visual vocabulary scored for every image.
A human-centered software portfolio.
A portfolio of human-centered software and design principles.
A human-centered software portfolio.
A portfolio of human-centered software and design principles.
A software portfolio.
A portfolio of software products and experimental projects.
Thoughtful interaction shaped around people, context, and practical use.
Thoughtful interaction shaped around the person and the situation, grounded in what they are actually trying to accomplish.
Thoughtful interaction shaped around people, context, and practical use.
Thoughtful interaction shaped around the person and the situation, grounded in what they are actually trying to accomplish.
Human-centered software with thoughtful interaction and practical use.
Human-centered software combining thoughtful interaction and practical use, with an emphasis on context-aware systems and adaptive behavior.
The Moments Collection
Helping people make everyday decisions with a growing body of tools.
Helping people make everyday decisions with a growing body of tools, each shaped around the situation to bring the most relevant options forward.
Helping people make everyday decisions with a growing body of tools.
Helping people make everyday decisions with a growing body of tools, each shaped around the situation to bring the most relevant options forward.
A growing body of context-aware tools for everyday decisions.
A growing body of context-aware tools for everyday decisions, combining dense underlying data with lightweight current signals to surface what matters most in the moment.
Prefer technical emphasis?
Prefer technical emphasis?
Currently viewing technical emphasis. Prefer general view?
Currently viewing technical emphasis. Prefer general view?
EXPLORATION 03
Visual preference matching
Which feels better to you?
Current selection
4 considerations
Didn’t find one? See what was ruled out 0
Static example: four matching images.
How this works
Step 1 — Record the signals
Start with a bounded set of images and record nine observable signals for every image.
How the signals are measured
- A
- Focal dominance
- B
- Contrast separation
- C
- Form weight
- D
- Element density
- E
- Directional energy
- F
- Rhythmic variation
- G
- Chroma
- H
- Palette separation
- I
- Color coverage
| Image | A | D | G | B | E | H | C | F | I |
|---|
Step 2 — Define visual attributes
Group related signals, then derive three continuous visual attributes from their weighted values.
Attribute formulas
Boldness
.40A + .35B + .25C
Activity
.40D + .35E + .25F
Color intensity
.45G + .30H + .25I
| Image | Boldness | Activity | Color intensity | Stimulation |
|---|
Step 3 — Interpret experiential patterns
Identify recurring visual patterns, normalize them as category scores, and test combinations that distinguish useful subsets of the image set.
Contextual tags
Readable choices combine several categories and appear only where they help distinguish nearby images.
- Open + Atmospheric → Let it breathe
- Kinetic + Structured + Graphic → Give it momentum
- Layered + Textural + Intricate → Bring out texture
Step 4 — Guide the choice
Present the richer model through a continuous target, a few contextual refinements, explicit limitations, and ranked alternatives.
Continuous target
Comfort ↔ Stimulate sets the initial multidimensional profile.
Adaptive refinements
The most useful tag recipes adjust category targets without becoming another full taxonomy.
Hard limitations
Ineligible image records are removed before scoring.
Ranked alternatives
The engine retains eligibility, total scores, components, and calculated preference effects.
Projects
PROJECT 01
Movie discovery
MomoMovies helps people choose a film by considering mood, viewing context, and available services.
MomoMovies helps people choose a film by weighing mood, viewing context, available services, and personal taste at the moment of choosing.
MomoMovies helps people choose a film by considering mood, viewing context, and available services.
MomoMovies helps people choose a film by weighing mood, viewing context, available services, and personal taste at the moment of choosing.
- Mood- and vibe-based film browsing.
- Filters for context, time, and services.
- Curated ordering for large catalogues.
- Private library and watch-intent models.
IN DEVELOPMENT
- Streaming-provider validation.
- Recommendation ranking evaluation.
- SvelteKit/TypeScript app on Vercel.
- Supabase auth, roles, library data, reviews, and poster storage.
- Generated catalog JSON with recommendation and Vibe-affinity data.
- TMDB metadata and U.S. streaming availability.
PROJECT 02
Music library tools
ListenLab is a private prototype for exploring music collections with more context.
ListenLab explores music collections through the relationships between songs, recordings, releases, playlists, and listening patterns.
ListenLab is a private prototype for exploring music collections with more context.
ListenLab explores music collections through the relationships between songs, recordings, releases, playlists, and listening patterns.
- Song, recording, release, and playlist relationships.
- Playlist organization and release-context prototypes.
- Collection views beyond standard streaming libraries.
- React/Vite frontend with FastAPI backend.
- Spotify OAuth/Web API for library, playlists, playback, follows, and listening data.
- SQLite migrations, encrypted tokens, history ingest, and identity records.
- Deduplication, caching, duration reconciliation, and track-family clustering.
PROJECT 03
Work context management
Momentum Management is a private prototype for seeing active work with more context.
Momentum Management gives active work more context across projects, tasks, inboxes, attention, and resumable state.
Momentum Management is a private prototype for seeing active work with more context.
Momentum Management gives active work more context across projects, tasks, inboxes, attention, and resumable state.
- SwiftUI dashboard for active work context.
- Selected product-root comparison.
- Inbox counts and local status summaries.
- Safe filesystem actions for configured roots.
IN DEVELOPMENT
- Single selected Productions root.
- Shared read-only operational views.
- Write actions gated by transaction rules.
- File-backed operations system for projects, tasks, status, inboxes, and context.
- Headless transactional commands for validated Codex actions.
- Confirmation tokens, journals, recovery checks, and schema validation.
- SwiftUI macOS 14 app over portable Markdown records.
Concepts
CONCEPT 01
Food decision aid
Meal Match
Meal Match treats choosing what to eat as a shared decision shaped by real constraints.
Choosing what to eat is often a coordination problem rather than a lack of options. Meal Match explores how preferences, energy, dietary constraints, time, and the people involved can shape a workable choice.
Meal Match is an early concept for making food decisions around real constraints.
Meal Match is an early concept for helping individuals or groups decide what to eat. It would bring preferences, context, and practical constraints together to move from indecision toward a workable choice.
Meal Match is early research into structured food-decision systems.
Meal Match is early product research into structured preference inputs, constraint matching, shared decisions, and recommendation logic. No application or functional prototype is available yet.
CONCEPT 02
Refined route navigation
PathWise
PathWise is an early idea about making route choices feel less demanding.
PathWise is an early concept for thinking about route choice in terms of the experience of getting there, not only the arrival time. Its questions and boundaries are still being defined.
PathWise is an early concept for more considered route choices.
PathWise is an early concept exploring how route choices might better reflect the conditions of a particular trip. Its identity, inputs, and product boundaries remain open.
PathWise is early research into context-aware route-choice systems.
PathWise is early product research into context-aware route choice. Its name, data assumptions, safety framing, and implementation boundaries remain open.
CONCEPT 03
Household chore coordination
Chore Thing
Chore Thing is an early idea about making household chores easier to choose and share.
Chore Thing is an early concept for making household chores easier to choose and share. Its questions around ownership, privacy, and collaboration are still open.
Chore Thing is an early concept for context-aware household chores.
Chore Thing is an early concept exploring how household chores might adapt to changing context. Its product identity and collaboration boundaries are still being defined.
Chore Thing is early research into context-adaptive household chore decisions.
Chore Thing is early research into context-adaptive household chore decisions. Its identity, privacy, ownership, and collaboration boundaries remain open.
Showcase
Selected visual work and experiments from across Kahntra.
EXPLORATION 01
Experience-based playlists
Part of ListenLab · View project
EXPLORATION 02
Film experience portraits
Part of MomoMovies · View project
