The Collective

Centerline Dynamic · Built for privacy
Version: 1.1.0 · Last updated July 7, 2026

Table of Contents

1. What Is The Collective?
2. Process Overview
3. The Synthesis Engine
4. Privacy & Data
5. Main Screen — Controls Reference
6. Multi-AI Panel Mode
7. Single-AI Thread Mode
8. History Screen
9. Trash Screen
10. Credits & Billing
11. Full Gesture / Interaction Reference

1. What Is The Collective?

The Collective lets you ask a question to multiple AI models simultaneously — Claude (Anthropic), Gemini (Google), ChatGPT (OpenAI), and Grok (xAI) — and synthesizes their answers into a consensus score and a summary. You can also run it as a one-on-one chat thread with a single AI or you can switch the AI model and tier with each response.

Your device never contacts AI providers directly, your questions are never stored on our servers, and no personal information is ever collected.

2. Process Overview

Multi-AI Mode (2+ providers selected)

When you ask a question to multiple AI providers, The Collective synthesizes their perspectives into a single unified summary using our proprietary refinement layer.

Depending on the number of rounds you select, the models will deliberate and review with increasing intensity:

• Level 1 (Fast): Independent AI responses.
• Level 2 (Deep): AI models review each other's work for higher accuracy.
• Level 3 (Ultimate): Iterative refinement for maximum consensus.

Once the deliberation is complete, all final-round answers are compared to generate an overall Agreement % and a per-AI score.

Key rule: The Collective's Synthesis Engine reads the original question and all AI answers. Your question format constraints (e.g. "summarize in 5 bullets per episode") override the synthesis engine's default style. The summary reflects the actual content of the answers, not a generic essay.

Single-AI Mode (1 provider selected)

You type a question
        │
        ▼
┌──────────────────────────────────────────┐
│  Single AI answers (no rounds/consensus) │
│  Conversation history is passed forward  │
│  so follow-up questions have context.    │
└──────────────────────────────────────────┘
        │
        ▼
   Thread continues until you tap
   the ↺ (New Question) button

3. The Synthesis Engine

The Collective's Synthesis Engine synthesizes multiple AI perspectives into a single unified summary using our proprietary refinement layer.

The summary does not truncate — even long structured responses (episode-by-episode TV breakdowns, multi-part analyses) complete fully.

4. Privacy & Data

Our philosophy

"AI should be private and optional. You shouldn't have to opt-out, you should have to opt-in."

What we collect

We collect no name, email, location, or any personally identifiable information.

Your questions

• Your device never contacts AI providers directly — our backend shields your identity.
• We do not store your questions on our servers.
• Questions you save to History are stored locally on your device only.
• AI providers may retain question content per their own policies (see links in the Privacy & Data Policy screen inside the app).

Standard Mode vs Live Search

Credits & payment

Credit purchases are processed entirely by Google Play. Centerline Dynamic, LLC never handles or stores payment information. Your balance is stored on our servers tied only to your anonymous device ID.

Your questions in History

History is stored in your device's local storage. Clearing app data or uninstalling the app will delete your local history. Nothing in your history is ever uploaded to our servers.

Transferring credits to a new device

Your balance is tied to your device ID, not your identity. To move credits:

1. Old phone: Long-press your credit balance → "Transfer code copied" toast → that code is now in your clipboard.
2. New phone: Tap your balance → "Restore balance from previous device" → paste the code → tap Transfer.
3. If the balance doesn't appear immediately, restart the app.

Keep your transfer code private — anyone with this code can claim your balance.

5. Main Screen — Controls Reference

Header Row 1

[ The Collective ]    [ 450 credits ]  [ ↺ ]  [ PDF ]  [ ⌚ ]

The About dialog

Tapping the app title opens the About dialog, which contains:

Header Row 2 — Settings (locked during an active thread)

[ Main ▾ ]  [ 2 rounds ▾ ]  [ 📶 ]  [ " ]

Settings in Row 2 and Row 3 are dimmed and locked while a query is running. In a single-AI thread they unlock between turns — you can switch the AI, tier, web search, and sources before each follow-up question. In a multi-AI session, to change settings first tap ↺ New Question.

Tier options

Rounds options

Note: When only 1 AI is selected, rounds are locked to 1. Multi-round deliberation only makes sense with multiple AIs reading each other.

Citation style (appears when Sources is enabled)

Header Row 3 — AI Provider chips

[ Claude ]  [ Gemini ]  [ ChatGPT ]  [ Grok ]

Selected AIs are highlighted with a colored background and border. Deselected AIs are dim/outlined.

6. Multi-AI Panel Mode

Active when 2 or more AIs are selected.

Question box

Appears below the header once a question is submitted.

[ Q  What are the best privacy tools for... ▾ ]

A 📶 wifi icon appears in the question box if Live Search was active for that query.

AI Panels

Each AI response is in its own card (panel). With multiple panels, they extend horizontally.

The panel header shows:

• AI provider name (colored)
• Specific model being used (e.g. claude-opus-4-8)

When multiple rounds are active, answers are stacked within the panel with "Round 1", "Round 2", "Round 3" labels.

Consensus + Summary sheet

After all AI answers arrive, a draggable sheet slides up from the bottom of the screen.

                    ───── (drag handle) ─────
         ┌──────────────────────────────────────┐
         │  Consensus: 84%  per-AI breakdown    │
         │                                      │
         │  Summary text...                     │
         │                                      │
         │  Sources (if enabled)                │
         │                                      │
         │         [ Export PDF ]               │
         └──────────────────────────────────────┘

Snap positions: 13% (peeking — you can still see panels), 45% (half screen), 100% (full screen).

The sheet never covers the question box — it always starts just below it.

Consensus panel

Shows:

• Overall agreement percentage (color-coded: red = low, green = high)
• Per-AI agreement scores

Summary panel

The single synthesized answer written by Gemini (or ChatGPT as fallback). Reads all AI answers + your original question. Format matches whatever constraints were in your question.

Shows actual credit cost used (e.g. "88 credits") at the bottom when complete.

Export PDF button

Available in the sheet (bottom) and in the header (top-right). Generates and shares a formatted PDF of the full session including question, all AI answers, consensus, and summary.

7. Single-AI Thread Mode

Active when exactly 1 AI is selected.

The question and answer appear as a chat thread. Each Q&A pair is its own tile. Follow-up questions continue the conversation with full context.

Thread tile

┌─────────────────────────────────────────┐
│  Q  What is quantum entanglement?       │
│ ─────────────────────────────────────── │
│  ● Claude                               │
│                                         │
│  Quantum entanglement is a phenomenon   │
│  where...                               │
│                                         │
│  ⊙ 12 credits                          │
└─────────────────────────────────────────┘

The credit count shows for the active turn when it completes. The cumulative thread cost is tracked for history.

Prior turns (collapsed above current)

Previous Q&A turns appear as collapsed tiles above the current one, capped at 35% of screen height with internal scroll.

Settings between turns

In a single-model AI thread:

• AI chips, tier, web search, and sources settings stay unlocked between turns — change them before each follow-up question. (Rounds stay locked to 1 in single-AI mode.)
• Each follow-up is answered by whichever AI and tier are selected when you send it, with the full conversation context carried forward — you can hand the same conversation from Claude to ChatGPT to Grok mid-thread.
• To start fresh with a clean slate, tap ↺ New Question in the header.

In a multi-model AI session:

• All AI chips, tier, rounds, web search, and sources settings are locked (dimmed) once the deliberation starts.
• This prevents mid-session configuration changes that would break the deliberation context.
• To start fresh with different settings, tap ↺ New Question in the header.

Thread history

The entire thread is saved as one history entry, titled by the first question. Each follow-up turn updates that same entry with the full transcript and cumulative credit cost. Tapping it in History reopens the full live thread — you can continue asking follow-ups.

8. History Screen

Access by tapping the ⌚ clock/history icon in the main header.

Shows your last N saved sessions (default: 250). Thread entries show as one card per conversation.

Session card

Each card shows:

• Question text (up to 2 lines)
• AI provider dots (colored)
• Session code: tier abbreviation, rounds, W (web search), C (citations)
• Date/time (relative: "5m ago", "2h ago", "3d ago", then M/D/Y)
• Consensus % badge (color-coded, multi-AI sessions only)
• 📶 wifi icon if live search was used
• 🔖 bookmark icon if archived
• PDF export button

Session codes:

• PT = Pro, MT = Main, EX = Express, WT = Who the...?
• 1R/2R/3R = rounds used
• W = live search was on
• C = citations/sources were on

Gestures on session cards

History screen header actions

9. Trash Screen

Deleted sessions live here for up to N days (default: 5) before permanent deletion.

10. Credits & Billing

How credits work

• Credits are purchased through Google Play in-app purchases.
• 1 credit ≈ 1 cent (USD). Example: 500 credits = ~$5.
• Credits are deducted based on actual tokens used, not the estimate.
• If an AI returns an error, you are not charged for that AI's response.
• Credits have no cash value and are non-refundable.

Credit estimate (the "Estimated credits · ..." line)

Shown above the text input. Tap it to see the detail dialog.

• Estimate is based on: tier × AI count × rounds × a scaling factor for question length.
• Estimate range: half to double the typical estimate (realistic spread for short vs. detailed answers).
• If your balance is below the estimate, the line turns red and reads "Not enough balance", and the Send button is disabled.
• Actual cost is shown on the completed session tile, not the estimate.

Low balance banner

A red banner appears when your balance drops below 50 credits. Tapping it opens the Top Up sheet.

Top Up sheet

Opened by tapping your balance chip.

Balance transfer dialog

Used to move credits from an old device to this one.

1. On your old device: long-press the balance chip → Transfer Code is copied to clipboard.
2. On your new device: Top Up → Restore balance → paste the code → Transfer.
3. Paste the entire code exactly as copied.
4. Restart the app if the balance doesn't update immediately.

11. Full Gesture / Interaction Reference

A single consolidated lookup table for every touch interaction in the app.

Main screen

Question box (multi-AI mode, while a session is active)

AI panels (multi-AI mode)

Consensus + Summary sheet (multi-AI mode)

Thread tiles (single-AI mode)

Prior turns (collapsed Q&A above current turn)

Query bar (bottom input area)

History screen

Trash screen

Top Up sheet

Balance transfer dialog

Appendix: Session Code Reference

Session codes appear in History cards in the format PT, 3R, W, C:

Appendix: Error Messages

The Collective is built by Centerline Dynamic, LLC. For support, visit centerlinedynamic.com