Keystone UX Simplification Plan — Make It Work for Normies

The Problem

Current Keystone chat requires users to understand:

"Skills" (what are they? when do I use them?)
"Σ₀" (no idea what this means)
"CSF Memory" (jargon)
"Convergence" (abstract concept)
"Context Modes" (overwhelming list)
Multiple AI providers (Auto, Claude, ChatGPT, Gemini, Grok, Local, Keystone FT)

Result: Non-technical users feel lost immediately. UI assumes programming knowledge.

Phase 1: First-Load Experience (Week 1)

On first visit, show:


"What would you like to do?"



[ 📝 Chat (Journal)      ] — Talk about your day

[ 💻 Code (Dev)          ] — Ask for code help

[ 🔍 Research (Explore)  ] — Learn something new

[ 💰 Trading (Trader)    ] — Market analysis

NO jargon. NO advanced options. Just four simple choices.

Implementation:

Check localStorage.lantern-hasSeenOnboarding
Show modal if false
Set flag to true on click
Route starter prompt based on choice

1.2 Provider Selection Simplified

Current: Dropdown withoptions New: Three radio buttons:


( ) Auto (I don't know — pick the best)

( ) Fast (Local, no API needed)

( ) Smart (Use my API key for best results)

Hide: All individual provider settings until user selects "Smart"

Phase 2: Remove Jargon (Week 1-2)

2.1 Rename UI Elements

Current	New	Why
CSF Memory / Symbols	Remember Important Things	Clear purpose
Context Modes	Give AI Context	Plain English
Convergence	—	Hide entirely from normies
Σ₀	—	Remove from UI
Skills	Quick Actions	More intuitive
Connectors	Tools	Familiar term
Performance Monitor	Show stats	Optional, clear intent

2.2 Add Help Bubbles

Every setting gets a ? icon with one-sentence explanation:


"AI Provider" ?

↓

"Chooses which AI assistant to use. Auto picks

 the best for your question. Smart uses your

 paid API keys for faster responses."

No multi-line explanations. One sentence max.

2.3 Hide Power User Settings

Add toggle: "Show Advanced Settings" in settings drawer

Basic view (default):

AI Provider (Auto / Fast / Smart)
Theme (Light / Dark)
Voice Mode (on/off)

Advanced view (when toggled):

Individual provider keys (Claude, Gemini, ChatGPT, Grok)
Context Modes (all checkboxes)
Connectors (MCP, Web Search)
Performance Monitor
Convergence Settings (for power users)

Phase 3: Intelligent Defaults (Week 2)

3.1 Auto-Detect What User Needs

Listen to first message, pick relevant context:


User says: "help me debug this" → Enable Code Context

User says: "should I buy..." → Enable Trading Context

User says: "what happened today" → Enable Web Search

User says: "remember..." → Enable Memory Mode

No user action needed. Context activates automatically based on intent.

3.2 Simplified Starter Prompts

Current:


⚡ What needs doing?

⚡ How's the system?

⚡ Get started

⚡ Tell me a story

⚡ Teach me

⚡ Play game

New (dynamic based on onboarding choice):

Chat mode:


⚡ What's on your mind?

⚡ Tell me about your day

⚡ Any questions?

Code mode:


⚡ Fix a bug for me

⚡ Explain this code

⚡ Design a feature

Research mode:


⚡ Teach me something

⚡ Explain how this works

⚡ Search for me

Trading mode:


⚡ Should I buy/sell?

⚡ Analyze this market

⚡ Trading ideas

Phase 4: Settings Restructure (Week 2-3)

Tabs:

General (Theme, Voice, Provider)
Connected Services (API keys, Tools)
Advanced (Context modes, Convergence, Monitoring)

Each tab = one concept. No overwhelming lists.

4.2 Progressive Disclosure

Show only what user needs:

First load: Just AI Provider + Theme
After first message: Show Voice, Context Mode (if relevant)
Aftermessages: Show Advanced if they ask

Use localStorage to track usage patterns.

Phase 5: One-Click Setup (Week 3)

5.1 Provider Setup Wizard

Instead of: "Enter your ANTHROPIC_API_KEY" Use:


"Get smarter responses"

  ↓

"Enter Claude API key" (or skip to use free option)

  ↓

[Paste key]

[Test key]

[Done]

Show a ✓ when key is valid.

5.2 Pre-Fill Common Actions

Instead of typing, offer:


Quick Actions:

☑ Chat with me

☑ Help with code

☑ Research

☑ Trade ideas

☑ Custom question

User clicks, AI responds without typing.

Phase 6: Feedback & Help (Week 4)

6.1 Contextual Tips

Show tips when user gets stuck:


No message for 2 min?

↓

💡 Tip: Try "Tell me about your day" or 

   ask me to explain something

6.2 Help Button

Single "?" button in nav → simplified help overlay:


What is Keystone?

  → An AI assistant that remembers you



How do I use it?

  → Just chat like normal. Ask for help.



What if I have a problem?

  → Click "Help" below, or email [email protected]

Phase 7: Mobile Optimization (Week 4)

7.1 Touch-Friendly Settings

Larger buttons (48px minimum)
Swipe-to-toggle instead of checkboxes
Bottom sheet drawer instead of modal

7.2 Reduced Settings on Mobile

Hide advanced settings entirely on phone. Only show:

AI Provider
Theme
Voice Mode

Access advanced via "More Settings" button.

Success Metrics

Metric	Target	How to Measure
First-time confusion	<2 min to first message	Analytics: time to first send
Provider selection	80% use "Auto"	localStorage preference logging
Help button clicks	<5% of users	Click tracking
Mobile UX rating	4.5+ stars	In-app feedback survey
Setup time	<1 minute	Timed user testing

Rollout Strategy

Week 1: Core UX

Onboarding modal
Simplified provider selection
Dark mode (done in #641)

Week 2: Jargon Removal + Defaults

Rename all elements
Help bubbles
Auto-detection
Dynamic prompts

Week 3: Settings Overhaul

Tab-based settings
Progressive disclosure
Wizard for API keys

Week 4: Polish + Mobile

Help system
Mobile optimization
User testing feedback

Design Principles

One choice at a time — Don't show all options upfront
Clear language — No jargon. No abbreviations (except common ones)
Smart defaults — Make the 80% case work without config
Progressive disclosure — Show advanced features only to power users
Mobile-first — Works great on phone, even better on desktop
Help nearby — Questions answered in the UI, not in docs
Quick setup — Working in <1 minute without reading docs

#641 — Dark mode (✅ DONE)
#642 — This epic (Keystone UX simplification)
Future: Onboarding modal
Future: Settings redesign
Future: Mobile optimization

Open Questions

Should "Skills" be hidden entirely for normies? (Yes, hide them)
Which AI provider should be default? (Claude, with fallback to Gemini)
How many starter prompts? (3-4 max per mode)
Should we track user behavior? (Yes, anonymously, for analytics)

Notes for Implementer

Use feature flags to roll out gradually
Test with non-technical user (ask a friend)
Measure time-to-first-message for each iteration
Get feedback before building the full wizard
Mobile testing is mandatory (use device emulator)

This is a user-first redesign. Every decision should ask: "Does a non-technical person understand this?"