# 07 — The Consensus Quiz MVP

> *A 10-question onboarding that produces a personalised CPML, seeds the consumer layer, and creates viral distribution for the entire protocol.*

## The product

At its simplest: answer 10 questions about your epistemic preferences, receive a map of your consensus profile (CPML), find people who share it, see news sources calibrated to it, and — daily — one question that refines your CPML slightly.

### Core loop

1. **Onboarding (5 minutes).** Ten questions designed to place a user along the major axes of epistemic preference — how much weight they give to scientific consensus, how they treat contested historical claims, which authority structures they recognise, how much uncertainty they prefer to see surfaced.

2. **Reveal.** The user sees their resulting CPML expressed both as a machine-readable profile (technical) and as a human-readable summary ("your consensus leans scientific-first with high uncertainty tolerance and strong secular framing; closest match: Pew Type 2 centre-left secular").

3. **Calibration refinement (ongoing).** One question per day, opt-in. Over 30 days, the profile becomes materially sharper. Over 90 days, it becomes an accurate model of the user's epistemic preferences.

4. **Discovery features.**
   - **Similar people** — show others with close CPMLs; optional anonymised community suggestion.
   - **Suggested sources** — news outlets, journals, podcasts whose editorial stance aligns with the user's CPML.
   - **Opposite view** — daily surfacing of "here is a claim you would strongly disagree with, and here is someone who believes it and why."
   - **Epistemic twin abroad** — someone in a distant culture whose CPML is closest to yours; what are they reading that you are not?

5. **Veritas integration (invisible to user at first).** Every time the user reads a claim in their browser (via the Veritas extension) or asks an AI agent something, their CPML composes verdicts from the plural-verdict substrate. The quiz is the onboarding; the protocol is the ongoing infrastructure.

## Why this is the right MVP

Three reasons it is strategically correct:

1. **It reaches general users directly.** The AI-labs-as-customer story (main Veritas revenue thesis) has a multi-year development horizon. The consumer quiz has a zero-day launch horizon. Starting with the quiz means starting with users, which is starting with real data, which is starting with real feedback for the design decisions that otherwise get made in a vacuum.

2. **It has viral mechanics.** Myers-Briggs-style self-reveal + shareable result + "this is me, in a way" + "share with your friends to compare" has proven viral for decades. 16Personalities reaches ~60 million users per year on exactly this pattern. A consensus profile quiz can plausibly reach similar numbers with the same mechanic, if executed well.

3. **It produces the product the protocol needs.** The CPML corpus — a real population of real users with real consensus preferences — is exactly what the plural-verdict substrate needs to function. Without CPMLs, the substrate is a database nobody queries correctly. The quiz generates the database of queriers.

## Design principles

These principles constrain the design to avoid harmful patterns.

### Principle 1 — Calibration, not cartoon

The 10 questions must place users on axes that are *real* (researched, defensible, not made up for viral effect). We will lose some viral mechanics by refusing to produce exaggerated, tribal-identity-flattering results. We will gain long-term legitimacy and avoid the "this quiz is basically a horoscope" critique that kills serious engagement.

### Principle 2 — Privacy-first by construction

CPMLs stay local by default. Browser or device storage. Cloud-sync is opt-in for cross-device users. Shared CPMLs are explicitly signed and published. No automatic upload, no hidden telemetry, no tracking the user's answers to build a profile that isn't theirs. This is also the GDPR-defensible posture.

### Principle 3 — Anti-echo-chamber is mandatory

If the product just clusters users with their tribe, it strengthens filter bubbles and the protocol becomes part of the problem. Mandatory counter-patterns:

- Default CPMLs set `surprising_opposite_view: daily` — the product surfaces high-quality content from frames the user would disagree with, every day, in a calibrated way.
- "Epistemic twin abroad" feature surfaces culturally distant but similar-frame users, expanding rather than contracting the user's world.
- Calibration questions regularly include "strong disagreement" framing ("here is a position the opposite end of your CPML would defend — can you steel-man it?").

### Principle 4 — Transparent scoring

The user sees exactly which axes the quiz evaluates, how their answers map, and why they were placed where they were placed. No hidden scoring, no manipulated results. This educates users about the framework and builds trust.

### Principle 5 — Opt-out is free

Privacy-maximal mode: use the protocol without a CPML. Veritas works (with foundation-reference-CPML defaults) for any user; the personalisation is an opt-in layer, not a requirement.

## Naming

Working name: **Compass** (as in "consensus compass"). Alternatives under evaluation:

- **Frame** — direct reference to the consensus-frame terminology.
- **Lens** — user views the world through their lens.
- **Prism** — a prism reveals components; the user sees the components of their consensus.
- **Mirror** — the product reflects the user's epistemic preferences to them.

Trademark and SEO check required before final selection. Early hypothesis: *Compass* is most evocative, matches the navigational metaphor, and has acceptable trademark space once narrowed to the specific category.

## Go-to-market

### Launch sequence

- **Week 1–8 Phase I MVP build.** Quiz engine + CPML generator + basic result view. No validator network yet; CPML is stored locally.
- **Week 8–12 Pre-launch.** Closed beta with a few hundred users. Iterate on question wording, result clarity, CPML accuracy.
- **Week 12–20 Soft launch.** Open to the public; link shared via Drow's / foundation's / working-group's networks. Minimal paid promotion.
- **Week 20–40 Viral expansion.** Seed through topical communities (journalism, academia, research, tech-policy, philosophy). Grow to 10–100K users.
- **Month 12+ Protocol integration.** Once Phase II of the main protocol ships, CPMLs populated during Phase I drive real queries against the validator network.

### Distribution

The quiz is a standalone product with a branded URL (e.g. `compass.veritas-protocol.org` or `veritas-compass.pages.dev`). Shareable result images auto-generated for each user ("My consensus profile: scientific-first, high uncertainty, low authority-deference. What's yours?"). Integration hooks for:

- Embeddable on partner sites (journalism, academic).
- API for apps that want to consume the CPML.
- Browser extension (later — reads page claims, applies CPML, surfaces verdicts).

## Critical analysis

**1 — The 10-question framework is a political object.** The axes chosen embed a theory of how consensus varies. Any choice of axes is contested. Response: publish the methodology publicly; invite critique; iterate on question wording with academic collaboration (moral foundations theory, political-compass research, values-test methodology). The framework is defensible if it is transparent and open to revision, not because it is neutral.

**2 — Early users are self-selected and unrepresentative.** The first 1,000 users will skew tech / academic / journalism / policy. Their CPMLs will not represent general population. Response: acknowledge explicitly; plan for broader-audience expansion in Week 20+; use stratified sampling to calibrate question wording for lay audiences.

**3 — Calibration bias in matchmaking.** If the product matches users with similar CPMLs, it creates filter bubbles. Principle 3 addresses this but cannot eliminate it. Response: ongoing measurement — track whether matched users' CPMLs converge or diverge over time; adjust matchmaking to favour divergence-preserving matches.

**4 — Data-privacy backlash if the product is perceived as surveillance.** Especially once browser-history-reading is introduced. Response: principle 2; explicit user consent per feature; raw data processing done client-side whenever possible; third-party privacy audit before scaling.

**5 — The viral mechanic attracts bad-faith appropriation.** Political-identity groups create "their" starter CPML that flatters their tribe; the product enables tribal identity-gaming rather than epistemic tooling. Response: third-party CPMLs are allowed and expected; the foundation doesn't police content; the UX surfaces how a CPML classifies the user relative to calibrated distributions, including opposing positions. Tribes that want to flatter themselves can, but the system's other features (opposite view, calibration) counter-balance.

**6 — Viral success overwhelms the protocol-backend that isn't ready yet.** If the quiz succeeds at 16Personalities scale but the validator network is still a prototype, users have a CPML that does nothing. Response: the quiz is useful standalone (matchmaking, source suggestions, opposite view) without the full substrate. Protocol integration adds value over time but is not required for launch utility.

**7 — Mis-categorisation harms.** A user is placed in a category they don't recognise; they feel stereotyped. Response: always present CPML as a *starting point*, not a definitive labelling; user can edit; the quiz result explicitly includes "ways we might have gotten this wrong."

## Related work (to be enriched by strategos + scout agents)

- **16Personalities, Political Compass, 8values, YourMorals.org** — viral-quiz mechanics and their critiques.
- **Spotify Wrapped, Ancestry onboarding** — personalised-reveal shareable-image patterns.
- **Pol.is / vTaiwan** — opinion-clustering methods that find consensus across opposed groups.
- **AllSides, Ground News** — multi-perspective news presentation.
- **Hypothes.is, Readwise** — annotation tools with community features.
- **OpenMind Platform, Heterodox Academy** — intellectual-humility training products.

## Open questions

- What is the ideal first set of 10 questions? This is the single most influential design decision.
- Should the product launch under Veritas branding, or as a standalone brand (`Compass`) with Veritas integration disclosed but not front-and-centre?
- Is browser-history reading a necessary feature, or a nice-to-have that adds too much privacy friction?
- Does the product monetise directly (subscriptions, premium features) or remain free to maximise reach, with Veritas revenue from other streams?
- What is the right partner cohort for early distribution — journalism, academia, tech, philosophy, policy — and how to avoid capture by any single one?

## What we'd build

- **`veritas-compass`** — standalone React / Next.js web app. Quiz engine + CPML generator + result renderer + daily-question feed + community discovery + opposite-view surfacer.
- **`cpml-starter-set`** — the first authoritative set of starter CPMLs referenced by the quiz.
- **`share-card-generator`** — automated image generation for shareable CPML-result cards.
- **`compass-api`** — API for apps that want to consume CPML data (read-only, user-authorised).
- **`compass-browser-ext`** — optional extension that applies CPML to pages read + surfaces verdicts from Veritas aggregators.
- **`compass-analytics`** — privacy-preserving aggregate analytics; publishes CPML distribution statistics, calibration, correlation with demographics if shared.
- **Published question-methodology document** — which axes, why, academic grounding, open to revision.