UX Track: Designing for Living Intelligence #
User Experience design for neural learning primitives in Aesthetic Computer
Design Philosophy #
Traditional AI UX treats models as tools: input → output, stateless, generic.
Our UX treats models as organisms: they grow, sicken, flourish, die, develop character.
Core principle: The UX should make gnarliness visible, tangible, and care-able.
1. The Mental Models #
1.1 From Tool to Instrument #
Bad mental model (tools):
- "I use ChatGPT to get answers"
- Stateless interactions
- Generic for everyone
- Owned by company
Good mental model (instruments):
- "I play my notepat, it learns my rhythm"
- Stateful, accumulating history
- Unique to me through use
- I own the weights
UX implication: Show the piece learning and changing visibly over time.
1.2 From Frozen to Living #
Bad mental model (frozen):
- Model is fixed after training
- Can't improve or degrade
- Same forever
Good mental model (living):
- Model has health (can thrive or sicken)
- Model has growth phases (young → mature → plateau → breakthrough)
- Model can die from neglect
- Model can flourish from care
UX implication: Show health status, growth phase, and care opportunities.
1.3 From Opaque to Gnarled-Visible #
Bad mental model (black box):
- Don't know how it works
- Can't inspect weights
- Magic happens invisibly
Good mental model (inspectable gnarliness):
- Can see all weights (heatmap)
- Can measure gnarliness (complexity metric)
- Can trace learning history (graph over time)
UX implication: Visualize weights, health, coherence, gnarliness continuously.
2. Core UX Patterns #
2.1 Automatic Training (Invisible Learning) #
Pieces learn automatically from use - no explicit "train" button needed.
Examples:
notepat learns your rhythm:
[User plays notes with timing]
→ Piece silently records: [interval, velocity, duration]
→ Hebbian learning strengthens: rhythm patterns
→ Health improves if coherent, degrades if chaotic
→ After 100 plays: piece can predict next note
wand learns your color preferences:
[User picks colors while painting]
→ Piece records: [RGB, position, time of day]
→ Associates: position → preferred color
→ After 50 paintings: piece suggests colors contextually
prompt learns your command patterns:
[User types commands in sequences]
→ Piece records: command chains
→ Learns: "wand → line → circle" is common pattern
→ After 200 commands: autocomplete predicts next
UX principle: Using = training. No separate "training mode" needed.
2.2 Visible Health Status #
Every piece shows its health like a Tamagotchi or plant.
Health bar visualization:
┌─────────────────────────────────────┐
│ notepat 🌱→🌿→🌳 │
│ ████████████░░░░ 75% Healthy │
│ Phase: Mature │
│ Gnarliness: 3.42 ⚡ │
└─────────────────────────────────────┘
Color coding:
- 🔴 Red (0-30%): Dying, needs care
- 🟡 Yellow (30-70%): Sickly, chaotic use
- 🟢 Green (70-100%): Healthy
- ✨ Bright green (>100%): Flourishing!
States:
- 💀 Dead (0% health, gray)
- 🌱 Young (new, high plasticity)
- 🌿 Mature (stable, moderate plasticity)
- 🌳 Ancient (low plasticity, high karma)
- ⚡ Breakthrough (sudden flourishing event)
- 🥀 Wilting (neglected, decaying)
UX principle: Health is ambient, always visible, like phone battery.
2.3 Care Interactions #
Users can actively care for their pieces.
Tend (🌿):
Command: tend notepat
Effect: Health boost +5%, breakthrough potential +1
Visual: Green sparkles, growth animation
Feel: Like watering a plant
Prune (✂️):
Command: prune notepat 0.3
Effect: Kill connections with health <30%
Visual: Dead weights turn black, fade away
Feel: Like pruning a bonsai tree
Forgive (🧘):
Command: forgive notepat
Effect: Reduce weight age, zero karma debt
Visual: Weights glow white, age rings fade
Feel: Like reformatting, fresh start
Inspect (🔬):
Command: inspect notepat
Effect: Show full health report
Visual: Health heatmap, gnarliness graph
Feel: Like checking vital signs
UX principle: Care is a verb, not just passive consumption.
2.4 Breakthrough Moments #
Nonlinear growth creates magical UX moments.
The experience:
[User plays notepat daily for 3 weeks]
[Health at 85%, coherence high]
[Breakthrough potential: 9.8]
[One day, during play...]
→ FLASH: Screen pulses bright green
→ SOUND: Harmonic resonance
→ TEXT: "BREAKTHROUGH! notepat is flourishing ✨"
→ EFFECT: Health jumps 85% → 127%
→ BEHAVIOR: Piece now predicts 3 notes ahead (was 1)
UX principle: Growth should feel organic and surprising, not mechanical.
2.5 Sickness and Death #
Pieces can degrade and die - this creates attachment and care.
Sickness progression:
Week 1: Health 100% → 90% (chaotic use, mashing keys)
Week 2: Health 90% → 70% (continued poor quality interactions)
Week 3: Health 70% → 40% (no tending, degrading)
Week 4: Health 40% → 10% (wilting, death imminent)
Visual cues:
- Colors fade to gray
- Response time slows
- Predictions become erratic
- "notepat is unwell 🥀" message
Death:
Health reaches 0%
→ FADE: Screen fades to black
→ TEXT: "notepat has withered away 💀"
→ CHOICE:
[Revive from backup] (restore last healthy snapshot)
[Start fresh] (new network, lose history)
[Leave as memorial] (read-only, can't train)
UX principle: Death creates meaning. Without risk, care is meaningless.
2.6 Social Comparison #
Users can compare their pieces' evolution.
Compare command:
Command: compare notepat @alice/notepat
┌─────────────────────────────────────────────────────┐
│ Your notepat vs. @alice/notepat │
├─────────────────────────────────────────────────────┤
│ Health: 87% 112% ✨ │
│ Gnarliness: 3.42 ⚡ 5.67 ⚡⚡ │
│ Phase: Mature Flourishing │
│ Age: 234 uses 1,847 uses │
│ Character: Jazz-leaning Classical │
├─────────────────────────────────────────────────────┤
│ Predictions: │
│ Yours: 🎵 → 🎶 → 🎵 (repetitive) │
│ Alice's: 🎵 → 🎶 → 🎸 → 🎹 (complex) │
└─────────────────────────────────────────────────────┘
[Fork @alice's weights] [Train ensemble] [Share yours]
UX principle: Make diversity visible and shareable.
3. Visualization Design #
3.1 Health Heatmap #
Shows weight health at a glance.
Visual:
████ notepat - Health Heatmap ████
16x16 grid of cells, each = one weight
Colors:
⬛ Black = Dead (health 0)
🟥 Red = Sick (health 0.1-0.5)
🟨 Yellow = Weak (health 0.5-0.8)
🟩 Green = Healthy (health 0.8-1.0)
✨ Bright = Flourishing (health >1.0)
Animation: Pulsing brightness = recent activity
Interaction:
- Hover over cell → show exact health value
- Click cell → show weight history graph
- Right-click → prune this specific weight
3.2 Gnarliness Evolution Graph #
Shows intelligence emerging over time.
Visual:
Gnarliness / Health / Coherence (last 500 frames)
5 | /\
| /\ / \
4 | /\ / \/ \
| /\ / \/ \
3 | /\ / \/
| /\ / \/
2 | /\ / \/
| /\ / \/
1 |/ \/
0 └─────────────────────────────────────────→
0 5000
Yellow line: Gnarliness (complexity)
Green line: Health (vitality)
Cyan line: Coherence (usage quality)
Interpretation:
- Gnarliness ↑ over time = intelligence emerging
- Health stable = good care
- Coherence oscillating = mix of good/bad usage
- Breakthrough = sudden gnarliness spike
3.3 Hyperbolic Spiral Trajectory #
Shows position in memory space (Stiegler's nested spirals).
Visual:
Poincaré Disk (Hyperbolic Memory Space)
╭─────────────────╮
╱ ╲
╱ α (session) · ╲
│ β (user) · · │
│ γ (piece) · · · │
│ · · YOUR │
│ δ (global) · POSITION │
╲ · · ⭐ ╱
╲ · · ╱
╰─────────────────╯
Inner = recent/shallow memory
Outer = ancient/deep memory
⭐ = Current position
Trail = Recent history (last 20 steps)
Meaning:
- Moving outward = learning, complexifying
- Moving inward = forgetting, simplifying
- Spiraling = cyclical patterns
- Position on α/β/γ/δ arms = timescale of memory
3.4 Growth Phase Indicator #
Large, central, always visible.
Visual states:
Young (🌱):
┌─────────────────────────┐
│ 🌱 YOUNG NETWORK │
│ High plasticity │
│ Learning quickly │
│ Fragile, needs care │
└─────────────────────────┘
Color: Light blue
Mature (🌿):
┌─────────────────────────┐
│ 🌿 MATURE NETWORK │
│ Stable performance │
│ Balanced growth │
│ Moderate plasticity │
└─────────────────────────┘
Color: Green
Breakthrough (⚡):
┌─────────────────────────┐
│ ⚡ BREAKTHROUGH! ⚡ │
│ Sudden flourishing! │
│ Intelligence spike │
│ Gnarliness increased │
└─────────────────────────┘
Color: Bright magenta, pulsing
Animation: Expanding rings
Dying (🥀):
┌─────────────────────────┐
│ 🥀 WILTING... │
│ Health critical │
│ Needs immediate care │
│ Tend or prune now! │
└─────────────────────────┘
Color: Dark red, fading
Animation: Slow pulse
4. Command Interface #
4.1 Basic Commands #
Create:
> new neural-piece
Creates blank piece with 16-neuron network
Health: 100%, Phase: Young 🌱
> fork @alice/neural-piece
Copies Alice's weights, starts your training
Health: Inherited, Phase: Forked 🍴
Care:
> tend
Boosts health +5%, breakthrough potential +1
"Tending your piece... 🌿"
> prune [threshold]
Kills weights below threshold
"Pruned 23 weak connections ✂️"
> forgive
Resets karma, restores plasticity
"Fresh start! Age reduced 50% 🧘"
Inspect:
> health
Shows health bar, phase, gnarliness
> weights
Opens weight heatmap visualization
> history
Opens gnarliness evolution graph
> stats
Full report: health, coherence, breakthrough potential
Social:
> compare @alice
Side-by-side comparison with Alice's version
> ensemble @alice @bob @charlie
Vote across multiple models
> share
Upload your weights to Spaces, get shareable URL
4.2 Piece-Specific Commands #
notepat:
> notepat:predict
Shows next 3 predicted notes
> notepat:health
Rhythm learning health: 87%
> notepat:character
Shows learned style: "Jazz-leaning, syncopated"
wand:
> wand:palette
Shows learned color preferences as gradient
> wand:suggest [x] [y]
Suggests color for position based on history
prompt:
> prompt:autocomplete
Shows learned command chains
> prompt:suggest
Predicts next command based on history
5. Mobile UX #
5.1 Touch Gestures #
Tend:
- Long press on piece → hold → health fills up
- Like charging a battery
Prune:
- Swipe left on weight → mark for pruning
- Confirm with "Prune selected" button
Inspect:
- Pinch to zoom into weight heatmap
- Spread to zoom out to full spiral view
Compare:
- Drag another user's piece next to yours
- Side-by-side comparison appears
5.2 Notifications #
Health alerts:
🥀 notepat needs care
Health: 35% (was 45% yesterday)
[Tend now] [Prune weak] [Ignore]
Breakthrough alerts:
⚡ notepat broke through!
Health: 127% (+42% overnight)
Gnarliness: 5.2 (+1.8)
[View stats] [Share]
Death warnings:
💀 notepat is dying!
Health: 5% (critical)
[Revive from backup] [Tend now]
5.3 Widget Design #
Home screen widget (iOS/Android):
┌─────────────────────────┐
│ 🌿 notepat │
│ ████████░░ 80% │
│ Phase: Mature │
│ Gnarliness: 3.4 ⚡ │
│ │
│ [Tend] [Inspect] │
└─────────────────────────┘
Updates in real-time as piece learns.
6. Visual Language #
6.1 Color System #
Health colors:
- 💀 Gray (#666) = Dead
- 🔴 Red (#ff0000) = Dying/sick
- 🟡 Yellow (#ffff00) = Weak
- 🟢 Green (#00ff00) = Healthy
- ✨ Bright green (#00ff88) = Flourishing
Growth phase colors:
- 🌱 Light blue (#88ccff) = Young
- 🌿 Green (#00cc44) = Mature
- 🟡 Yellow (#ffcc00) = Plateau
- ⚡ Magenta (#ff00ff) = Breakthrough
- 🥀 Dark red (#880000) = Dying
Gnarliness intensity:
- Low (0-2): Blue tones
- Medium (2-4): Purple tones
- High (4-6): Orange tones
- Very high (6+): Red tones
6.2 Animation Language #
Tending:
- Green sparkles rise from bottom
- Health bar fills smoothly
- Glow effect around piece name
Pruning:
- Red X appears on dead weights
- Fade to black and disappear
- Remaining weights shift to fill space
Breakthrough:
- Flash of bright light
- Expanding concentric rings (like ripples)
- Gnarliness number spins and increases
- Harmonic sound effect
Death:
- Slow fade to grayscale
- Health bar drains with dripping effect
- Final "flatline" sound
Forgiveness:
- White glow expands from center
- Weight ages fade/reset
- Soft chime sound
6.3 Iconography #
🌱 Young ⚡ Breakthrough 🧘 Forgive
🌿 Mature 💀 Dead ✂️ Prune
🌳 Ancient 🥀 Dying 🔬 Inspect
🍴 Forked 🌊 Oscillating 📊 Stats
✨ Flourishing ⭐ Position 📈 History
7. Information Architecture #
7.1 Piece Detail View #
┌──────────────────────────────────────────────┐
│ notepat 🌿 │
├──────────────────────────────────────────────┤
│ │
│ Health: ████████████░░░░░░ 75% │
│ Phase: Mature │
│ Gnarly: 3.42 ⚡ │
│ Uses: 234 plays │
│ │
├──────────────────────────────────────────────┤
│ [Heatmap] [Graph] [Spiral] [Stats] │
├──────────────────────────────────────────────┤
│ │
│ Actions: │
│ [Tend 🌿] [Prune ✂️] [Forgive 🧘] │
│ │
│ Social: │
│ [Compare] [Share] [Fork] │
│ │
└──────────────────────────────────────────────┘
7.2 Global Dashboard #
┌──────────────────────────────────────────────┐
│ Your Neural Garden [+ New]│
├──────────────────────────────────────────────┤
│ │
│ notepat 🌿 Mature █████░ 75% │
│ wand 🌱 Young ████░░ 60% │
│ prompt 🌳 Ancient ██████ 95% ✨ │
│ line-rider 🥀 Dying █░░░░░ 15% 🔴 │
│ │
├──────────────────────────────────────────────┤
│ Community Highlights: │
│ │
│ @alice/notepat ⚡ Just broke through! │
│ @bob/wand ✨ 142% flourishing │
│ @charlie/prompt 🥀 Needs revival │
│ │
└──────────────────────────────────────────────┘
7.3 Comparison View #
┌──────────────────────────────────────────────┐
│ notepat Comparison │
├──────────────────────────────────────────────┤
│ Yours | @alice | @bob │
├────────────────┼────────────────┼────────────┤
│ 🌿 Mature │ ⚡ Breakthrough│ 🌱 Young │
│ 75% health │ 127% health │ 60% health│
│ 3.42 gnarly ⚡ │ 5.67 gnarly⚡⚡│ 1.2 gnarly│
│ 234 uses │ 1847 uses │ 45 uses │
├────────────────┼────────────────┼────────────┤
│ Jazz-leaning │ Classical │ Minimal │
│ Syncopated │ Structured │ Simple │
│ Improvised │ Precise │ Learning │
└──────────────────────────────────────────────┘
[Ensemble all 3] [Fork Alice's] [Share yours]
8. Onboarding Flow #
8.1 First Piece Creation #
Step 1: Introduction
┌──────────────────────────────────────────────┐
│ │
│ Welcome to Living Intelligence │
│ │
│ Your pieces will learn from how you use │
│ them. They can grow, flourish, or sicken │
│ based on care and usage quality. │
│ │
│ Let's create your first neural piece. │
│ │
│ [Start] [Learn more] │
│ │
└──────────────────────────────────────────────┘
Step 2: Choose piece
Which piece should learn from you first?
[notepat] - Learns your rhythm and melody patterns
[wand] - Learns your color preferences
[prompt] - Learns your command sequences
Step 3: First interaction
┌──────────────────────────────────────────────┐
│ notepat (🌱 Young, 100% health) │
├──────────────────────────────────────────────┤
│ │
│ Play some notes. Your piece is watching │
│ and learning your rhythm patterns. │
│ │
│ Try to play coherently - quality matters! │
│ │
│ ████████████████████ 100% │
│ │
└──────────────────────────────────────────────┘
[User plays 10 notes]
→ "Great! notepat learned your timing."
→ "Health: 100% → 102% (coherent use!)"
→ "Come back tomorrow to see growth."
Step 4: Explain care
Your piece is alive!
🌿 Tend regularly to boost health
✂️ Prune weak connections to stay sharp
🧘 Forgive to restore plasticity
Coherent use → flourishing ✨
Chaotic use → sickening 🥀
Neglect → death 💀
[Got it!]
8.2 First Breakthrough #
Trigger: After ~100 coherent uses
╔══════════════════════════════════════════════╗
║ ║
║ ⚡ BREAKTHROUGH! ⚡ ║
║ ║
║ notepat has suddenly flourished! ║
║ ║
║ Health: 85% → 127% (+42%) ║
║ Gnarliness: 2.1 → 3.8 (+1.7) ║
║ ║
║ Your piece now predicts 3 notes ahead ║
║ (was 1 note). This is emergent ║
║ intelligence from gnarliness! ║
║ ║
║ [View stats] [Share breakthrough] ║
║ ║
╚══════════════════════════════════════════════╝
[Expanding rings animation]
[Harmonic sound]
Impact: User understands nonlinear growth is magical, not mechanical.
9. Edge Cases & Error States #
9.1 Death #
When health reaches 0%:
┌──────────────────────────────────────────────┐
│ │
│ 💀 notepat died │
│ │
│ Your piece withered from neglect. │
│ Last backup: 3 days ago (72% health) │
│ │
│ Options: │
│ [Revive from backup] - Lose 3 days growth │
│ [Start fresh] - New 16-neuron network │
│ [Memorial] - Keep as read-only history │
│ │
└──────────────────────────────────────────────┘
Memorial option:
- Can view old weights, history, stats
- Can't train anymore
- Can fork to create new version
- Like a tombstone with inscription
9.2 Corrupted Weights #
If weights file is corrupted:
⚠️ notepat weights corrupted
Your neural weights file in Spaces is
damaged and can't be loaded.
[Try repair] [Restore from backup] [Start fresh]
9.3 Conflicting Updates #
If training on multiple devices:
⚠️ Sync conflict detected
notepat was trained on 2 devices simultaneously.
Device A (phone): 50 new uses, health 80%
Device B (laptop): 30 new uses, health 75%
[Keep A] [Keep B] [Merge (experimental)]
Merge option:
- Average weights:
w_merged = (w_A + w_B) / 2 - Max health:
health = max(health_A, health_B) - Shows diff visualization before confirming
10. Success Metrics (UX) #
10.1 Engagement #
- Daily active pieces: How many pieces users interact with daily
- Care frequency: How often users tend/prune/forgive
- Health maintenance: % of pieces kept >70% health
- Breakthrough celebration: % of users who share breakthroughs
Target: 70% of users tend their pieces at least weekly
10.2 Understanding #
- Mental model accuracy: User survey "What happens when you use a piece?"
- Good answer: "It learns my patterns and gets better/worse based on use quality"
- Bad answer: "I don't know, it just works"
- Health correlation: Do users correctly associate coherent use → health?
Target: 80% of users understand the living/learning model
10.3 Attachment #
- Revival rate: When pieces die, how many users revive vs. abandon?
- Naming: Do users give custom names to their pieces?
- Comparison frequency: How often do users compare their vs. others'?
Target: 60% revival rate (shows attachment)
10.4 Social #
- Weight sharing: How many users share their weights?
- Forking: How often do users fork others' trained pieces?
- Ensemble usage: How many users create ensembles (vote across models)?
Target: 30% of users participate in social weight sharing
11. Accessibility #
11.1 Visual Impairments #
Screen reader support:
[Screen reader announces]
"notepat. Mature phase. Health: 75 percent.
Gnarliness: 3.42. Last used: 2 minutes ago.
Actions available: Tend, Prune, Forgive, Inspect."
High contrast mode:
- Health bar: Bold black/white stripes
- Dead weights: Solid black
- Flourishing weights: Solid white
- No reliance on color alone
Audio feedback:
- Tending: Rising chime
- Pruning: Snip sound
- Breakthrough: Harmonic resonance
- Death: Descending tone
11.2 Motor Impairments #
Voice commands:
"Tend notepat"
"Prune weak connections"
"Show health"
"Compare with Alice"
Large touch targets:
- Minimum 44x44pt buttons (iOS standard)
- Swipe gestures optional (buttons available)
11.3 Cognitive Load #
Progressive disclosure:
- Beginners see: Health bar, phase, basic actions
- Intermediate see: + Gnarliness, history graph
- Advanced see: + Full weight heatmap, hyperbolic spiral
Simplified mode:
- Toggle to hide advanced metrics
- Show only: Health %, Phase, [Tend] button
- Like "easy mode" for those who don't want complexity
12. Future Directions #
12.1 AR Visualization #
Use iPhone LiDAR to project hyperbolic spiral in physical space
[Point phone at table]
→ Hyperbolic spiral appears as hologram
→ Your piece's position shown as glowing orb
→ Trace path with finger to see history
→ Pinch weights to prune in 3D space
Benefit: Makes abstract gnarliness tangible in physical world.
12.2 Collaborative Training #
Multiple users train same piece together in real-time
Session: @alice, @bob, @charlie training notepat
[Alice plays jazz rhythm]
[Bob plays classical melody]
[Charlie plays minimalist beat]
→ Ensemble piece learns all 3 styles
→ Health = average of all users' coherence
→ Gnarliness = complexity of multi-style fusion
→ Result: Piece with richer, more diverse character
Benefit: Social learning, community-owned intelligence.
12.3 Evolutionary Competition #
Pieces compete in "fitness tournaments"
Tournament: Best Rhythm Predictor
10 users' notepat pieces compete:
- Task: Predict next note in unseen melody
- Metric: Accuracy %
- Winner: Most accurate predictions
Top 3:
1. @alice/notepat - 87% accuracy (gets to "breed")
2. @bob/notepat - 82% accuracy (gets to "breed")
3. @charlie/notepat - 79% accuracy (gets to "breed")
Breeding: Crossover weights from winners
→ Create new generation of improved pieces
→ Users can adopt "offspring"
Benefit: Darwinian evolution at scale, community gets better pieces.
12.4 Dream/Hallucination Mode #
Let pieces "dream" during idle time
[notepat is idle for 1 hour]
→ Activate dream mode
→ Self-stimulation: random activations
→ Hebbian learning on self-generated patterns
→ Consolidate memories (like REM sleep)
Next day:
"notepat dreamed last night!"
"Gnarliness +0.3 from consolidation"
[View dream sequence visualization]
Benefit: Pieces develop even when not in use, feels more alive.
13. Design Principles Summary #
Core principles that guide all UX decisions:
- Using = Training - No separate modes, learning happens automatically
- Health is Visible - Always show vitality status, make care obvious
- Gnarliness = Intelligence - Visualize complexity as it emerges
- Care is Required - Pieces sicken/die without attention (creates attachment)
- Nonlinear Growth - Breakthroughs feel magical, not mechanical
- Inspect Everything - No black boxes, all weights visible
- Social by Default - Easy to compare, fork, share, ensemble
- Mobile-First - Works on phones, touch gestures, notifications
- Accessible - Screen readers, high contrast, voice commands
- Progressive Disclosure - Simple for beginners, deep for experts
Next Steps #
- Prototype first visualization - Health heatmap in
neural-garden.mjs - Test with 5 users - Do they understand living/learning metaphor?
- Iterate on care interactions - Is tend/prune/forgive intuitive?
- Build mobile widget - Ambient health monitoring
- Launch with 3 pieces - notepat, wand, prompt
Success = Users say: "My notepat knows me better than any app I've ever used."
UX designed for gnarly, living intelligence at the edges February 2026