AI Model Discovery Timeline: User Guide¶

Table of Contents¶

What is the Discovery Timeline?
Getting Started
Navigating the Timeline
Using Filters
Understanding Event Cards
Community Insights
FAQ
Troubleshooting
Providing Feedback

What is the Discovery Timeline?¶

The AI Model Discovery Timeline is MAGIEVA's real-time feed of AI model discoveries, pricing changes, and capability updates from major providers like OpenAI, Anthropic, Google, AWS, and more.

Why use it? - Stay informed: Never miss important AI model updates - Compare providers: See pricing changes across providers in one place - Plan ahead: Know when models are deprecated or updated - Learn from community: Read tips, tricks, and best practices

Example events: - OpenAI launches GPT-5 (model_discovered) - Anthropic reduces Claude 3.5 pricing by 20% (pricing_change) - Google Gemini adds image generation (capability_update) - AWS Bedrock deprecates Titan Text v1 (model_deprecated)

Getting Started¶

Accessing the Discovery Timeline¶

Log in to MAGIEVA: https://app.magieva.com
Click "Discovery" in the left sidebar (between "Conversations" and "Integrations")
You'll see the timeline with the latest AI model events

First-time users: The timeline may be empty if no discovery has run yet. Check back after the hourly discovery job completes.

System Requirements¶

Browser: Chrome 120+, Firefox 120+, Safari 17+, Edge 120+
Internet: Stable connection (timeline events load via infinite scroll)
Permissions: Standard user account (no special permissions required)

Navigating the Timeline¶

Timeline Layout¶

The Discovery Timeline has three main sections:

┌─────────────────────────────────────────────────┐
│  Header (Title + Description)                   │
├─────────────────────────────────────────────────┤
│  Filters (Provider, Event Type, Search)         │
├─────────────────────────────────────────────────┤
│                                                  │
│  Event Cards (infinite scroll)                  │
│  ┌────────────────────────────────────────┐    │
│  │ [Provider Icon] OpenAI                 │    │
│  │ Model Discovered • 2 hours ago         │    │
│  │ GPT-5 launched with 10M context window │    │
│  └────────────────────────────────────────┘    │
│  ┌────────────────────────────────────────┐    │
│  │ [Provider Icon] Anthropic              │    │
│  │ Pricing Change • 5 hours ago           │    │
│  │ Claude 3.5 Sonnet reduced to $3/1M tok │    │
│  └────────────────────────────────────────┘    │
│                                                  │
│  [Load More automatically as you scroll]        │
└─────────────────────────────────────────────────┘

Infinite Scroll¶

Events load automatically as you scroll down: - Initial load: 20 events - Auto-load: Next 20 events when you scroll near the bottom - Total: Unlimited (all historical events available)

Tip: Scroll slowly for smooth loading. Fast scrolling may trigger multiple loads at once.

Using Filters¶

Filters help you narrow down the timeline to specific providers, event types, or search queries.

Provider Filter¶

What it does: Show only events from a specific AI provider

How to use: 1. Click the Provider dropdown (default: "All Providers") 2. Select a provider from the list: - OpenAI - Anthropic - Google - AWS (Bedrock) - Azure (OpenAI Service) - Cohere - Mistral - AI21 Labs - Replicate - Hugging Face 3. Timeline updates automatically

Example: Select "Anthropic" to see only Claude model updates

Event Type Filter¶

What it does: Show only specific types of events

How to use: 1. Click the Event Type dropdown (default: "All Events") 2. Select an event type: - Model Discovered: New AI models launched - Pricing Change: Price increases/decreases - Capability Update: New features, modalities, or capabilities - Model Deprecated: Models being retired - API Version Change: API breaking changes or new versions 3. Timeline updates automatically

Example: Select "Pricing Change" to see only pricing updates

What it does: Search across model names, providers, and event descriptions

How to use: 1. Click the Search input box 2. Type your query (e.g., "GPT-4", "image generation", "price") 3. Search executes automatically after 300ms (debounced)

Search tips: - Model names: "GPT-4", "Claude 3.5", "Gemini Pro" - Capabilities: "vision", "audio", "multimodal" - Pricing terms: "free tier", "discount", "price increase" - Providers: "OpenAI", "Google", "Anthropic"

Example: Search "vision" to find all models with vision capabilities

Combining Filters¶

You can use multiple filters together:

Example 1: Provider = "OpenAI" + Event Type = "Pricing Change" - Result: Only OpenAI pricing changes

Example 2: Search = "audio" + Event Type = "Capability Update" - Result: Only capability updates related to audio

Clear filters: Click "Clear All" button or reload the page

Understanding Event Cards¶

Each event card shows key information about the AI model update.

Event Card Anatomy¶

┌──────────────────────────────────────────────────┐
│ [Provider Logo] Provider Name            [Badge] │  ← Header
│ Event Type • Timestamp                           │  ← Metadata
│ ─────────────────────────────────────────────── │
│ Event Title (e.g., "GPT-5 launched")            │  ← Title
│ Event Description (details about the change)     │  ← Description
│ ─────────────────────────────────────────────── │
│ Changes:                                         │  ← Change Details
│ • Price: $10/1M → $8/1M tokens (-20%)           │
│ • Context Window: 128k → 200k tokens            │
│ ─────────────────────────────────────────────── │
│ [View Model Details →]                          │  ← Actions
└──────────────────────────────────────────────────┘

Event Types & Badges¶

Each event type has a color-coded badge:

🔵 Model Discovered (Blue) - New model launched
🟢 Pricing Change (Green) - Price decreased
🟠 Pricing Change (Orange) - Price increased
🟣 Capability Update (Purple) - New features added
🔴 Model Deprecated (Red) - Model being retired
🟡 API Version Change (Yellow) - API breaking changes

Timestamps¶

Timestamps show how long ago the event occurred: - "2 minutes ago" - Very recent - "3 hours ago" - Today - "2 days ago" - This week - "Jan 15, 2026" - Older events (absolute date)

Tip: Events are sorted newest-first. The most recent event is always at the top.

Change Details¶

Pricing and capability changes include specific details:

Pricing Change Example:

Changes:
• Input Price: $3/1M tokens → $2.5/1M tokens (-16.7%)
• Output Price: $15/1M tokens → $12/1M tokens (-20%)
• Context Window: 200k tokens (unchanged)

Capability Update Example:

Changes:
• Added: Vision (image understanding)
• Added: Function Calling (native tool use)
• Context Window: 128k → 200k tokens

Actions¶

Each event card has action buttons:

"View Model Details →": Opens the full model page with documentation
"See Insights": View community tips and best practices (if available)

Community Insights¶

Community Insights are tips, tricks, and best practices shared by MAGIEVA users about specific AI models.

Viewing Insights¶

Where to find them: 1. Click "View Model Details" on any event card 2. Scroll to the "Community Insights" section 3. Browse insights by type: - Best Practices: Recommended usage patterns - Code Examples: Working code snippets - Limitations: Known issues and workarounds - Usage Guides: Step-by-step tutorials

Insight Card Anatomy¶

┌──────────────────────────────────────────────────┐
│ [Icon] Best Practices                     [Badge]│
│                                                   │
│ Title: "Use Claude 3.5 for code generation"      │
│ Description: "Claude 3.5 Sonnet excels at..."    │
│                                                   │
│ 👍 42 votes   💬 8 comments                      │
│ [View Source →]                                   │
└──────────────────────────────────────────────────┘

Suggesting an Insight (Coming in Sprint 3)¶

Currently, insights are read-only. In Sprint 3, you'll be able to: - Submit your own tips and tricks - Vote on helpful insights - Comment on insights - Earn reputation points

Interested? Click "Suggest an Insight" to join the waitlist and be notified when submissions open.

FAQ¶

General Questions¶

Q: How often does the Discovery Timeline update? A: The discovery job runs every hour, checking for new models, pricing changes, and capability updates across all providers.

Q: How far back does the timeline go? A: The timeline shows all events since the discovery service launched (2026-03-13). Historical data before this date is not available.

Q: Can I get notifications for specific events? A: Not yet! Event notifications (email, Slack, WebSocket) are planned for Sprint 3.

Q: Is the timeline real-time? A: Yes and no. The discovery job runs hourly, so events appear within 1 hour of being announced by the provider. WebSocket updates for instant notifications are coming in Sprint 3.

Pricing Questions¶

Q: Why do I see a "pricing change" for a model I don't use? A: The Discovery Timeline shows all events from all providers, regardless of your current usage. Use filters to narrow down to providers you care about.

Q: Are pricing changes retroactive? A: No. Pricing changes apply to new API calls after the change date. Your existing usage is billed at the old rate until the change takes effect.

Q: How accurate are the pricing comparisons? A: Pricing data is fetched directly from provider APIs and documentation. However, pricing can vary by region, volume discounts, and enterprise agreements. Always verify with the provider's official pricing page.

Technical Questions¶

Q: Why do some event cards show "No changes detected"? A: This occurs when the discovery job detects a new model but the provider hasn't published detailed specifications yet. The card will update once more information is available.

Q: What does "Rate limit exceeded" mean? A: The Discovery Timeline API has a rate limit of 60 requests per minute per user. If you see this error, wait 1 minute before refreshing or scrolling for more events.

Q: Why is infinite scroll not loading more events? A: Possible reasons: 1. You've reached the end of the timeline (no more events) 2. Network connection issue (check your internet) 3. Rate limit exceeded (wait 1 minute) 4. Browser issue (try refreshing the page)

Q: Can I export the timeline to CSV or JSON? A: Not yet! Export functionality is planned for Sprint 3. For now, you can manually copy event details or take screenshots.

Troubleshooting¶

Issue: Page won't load (blank screen)¶

Symptoms: Discovery page shows blank screen or loading spinner indefinitely

Possible Causes: - Discovery service is down - Database migration not applied - Network connection issue

Solutions: 1. Refresh the page (Cmd+R or Ctrl+R) 2. Clear browser cache and reload 3. Try a different browser 4. If the issue persists, contact support@magieva.com

Issue: Timeline events not displaying¶

Symptoms: Page loads but shows "No events found" or empty timeline

Possible Causes: - Discovery job hasn't run yet (first-time setup) - Filters too restrictive (no events match) - Database connection issue

Solutions: 1. Clear all filters and try again 2. Check back in 1 hour (discovery job runs hourly) 3. Try searching for a specific model (e.g., "GPT-4") 4. Report via #discovery-beta if issue persists

Issue: Filters not working¶

Symptoms: Selecting a filter doesn't update the timeline

Possible Causes: - Browser JavaScript disabled - React Query cache issue - API endpoint error

Solutions: 1. Refresh the page 2. Clear browser cache 3. Check browser console for errors (F12 → Console) 4. Try using incognito/private browsing mode 5. Report via #discovery-beta with browser details

Issue: Infinite scroll not loading more events¶

Symptoms: Scrolling to bottom doesn't load additional events

Possible Causes: - Reached end of timeline (no more events) - Rate limit exceeded (too many requests) - Network timeout

Solutions: 1. Check if "Load More" indicator is visible (should appear before scrolling) 2. Wait 1 minute (rate limit reset) 3. Refresh the page 4. Try scrolling more slowly 5. Report via #discovery-beta if issue persists

Issue: "Rate limit exceeded" error (429)¶

Symptoms: Red toast notification: "Rate limit exceeded. Please try again later."

Cause: You've made more than 60 API requests in 1 minute (usually from rapid scrolling or refreshing)

Solution: 1. Wait 60 seconds before trying again 2. Scroll more slowly (avoid fast scrolling) 3. Avoid refreshing the page repeatedly 4. If you need high-volume access, contact engineering to increase your limit

Providing Feedback¶

We want to hear from you! Your feedback helps us improve the Discovery Timeline.

Beta Feedback Form¶

Email: support@magieva.com with the subject "Discovery Timeline Feedback"

What to include: - What you liked - What confused you - Bugs you encountered - Features you wish existed

Slack Channel¶

Join the #discovery-beta Slack channel for: - Quick questions - Bug reports - Feature requests - Community discussions

Email Support¶

Email: support@magieva.com Subject: "Discovery Timeline Feedback"

What's Next?¶

Sprint 3 Roadmap (April 2026)¶

Planned Features: - ✨ Event Notifications: Email, Slack, and in-app alerts for new events - 📊 Advanced Filters: Date range picker, multi-provider selection - 💬 Community Contributions: Submit and vote on insights - 📈 Analytics Dashboard: Track model adoption, pricing trends - 🌍 3D Model Universe: Interactive visualization of AI model landscape - 📁 Export to CSV/JSON: Download timeline data for analysis

Public Launch (Q2 2026)¶

The Discovery Timeline will be available to all MAGIEVA users (not just internal) by June 2026. We'll also add: - Provider health dashboard (API uptime, latency) - Model comparison tool (side-by-side pricing, capabilities) - Custom alerts (e.g., "notify me when GPT-5 price drops")

Appendix¶

Supported Providers¶

The Discovery Timeline currently tracks 15 AI providers:

OpenAI (GPT-4, GPT-3.5, DALL-E, Whisper)
Anthropic (Claude 3 family)
Google (Gemini, PaLM 2)
AWS Bedrock (Titan, Claude, Llama via AWS)
Azure OpenAI Service (GPT-4, GPT-3.5 via Azure)
Cohere (Command, Embed)
Mistral AI (Mistral Large, Medium, Small)
AI21 Labs (Jurassic-2)
Replicate (Open-source models)
Hugging Face (Open-source models)
Stability AI (Stable Diffusion)
Meta (Llama 2, Llama 3 via Replicate/Hugging Face)
Perplexity AI (pplx models)
Together AI (Open-source models)
Fireworks AI (Open-source models)

Coming soon: Inflection, Writer, Writer, and more!

Glossary¶

Model: An AI model (e.g., GPT-4, Claude 3.5 Sonnet)
Provider: The company offering the AI model (e.g., OpenAI, Anthropic)
Event: A detected change (discovery, pricing change, capability update, etc.)
Timeline: Chronological list of events
Insight: Community-submitted tip, trick, or best practice
Discovery Job: Automated process that checks provider APIs for changes (runs hourly)
Rate Limit: Maximum number of API requests allowed per time period (60/min)

Questions? Contact us at support@magieva.com or Slack #discovery-beta

Last Updated: 2026-03-13 Version: 1.0 (Sprint 2 Week 4)