Preview Feedback Loop

Last updated 2026-07-02. This is the operating path for showing Values Commons to a small circle without turning it into a public launch.

Goal

The next milestone is not more surface area. It is learning whether real people can use the current system for one real decision, trust the evidence, and tell us what blocked them.

Use this loop for grant reviewers, trusted testers, and the first 20-50 people you can invite personally.

Before Sending A Link

Build the private preview and run the gates:

npm run build:preview
npm run audit:preview
npm run verify:full

The preview should be noindexed, static, and self-contained. Send the deployed root URL, not raw repository files.

Who To Invite

Invite people who can test one concrete wedge:

Someone considering a bank, credit union, or divestment move.
Someone choosing a digital tool and worried about privacy.
Someone who buys clothing, electronics, or learning resources and cares about repair, labor, or openness.
One or two technically curious people who can inspect sources and tell you where trust breaks.

Do not ask for general opinions first. Ask them to make or evaluate one real decision.

Three Test Tracks

Track	Ask them to test	Evidence to collect
Individual choice	Use a live instance for one real decision, then inspect the evidence behind one result.	Did the ranking change when their values changed? Which source or option affected trust?
Group decision	Take a small group through Assembly, Workshop, and Slate for a real shared choice.	Where did alignment appear, where did dissent remain, and what blocked commitment?
Instance stewardship	Use the Adoption Kit to scope a possible domain front door before code.	Was the decision clear, was the starter lens plausible, and was the contribution route visible?

The Tester Task

Ask each tester to choose one track, then do this:

Open the preview root and take the two-minute tour.
For an individual choice, open a live instance, set real values, inspect at least one result, and try one category such as /app/#explore/banking, /app/#explore/ai-assistants, /app/#explore/clothing, or /app/#explore/learning-resources.
For a group choice, open Assembly and Slate, then name both agreement and dissent.
For instance stewardship, open the Adoption Kit and fill the steward packet before touching code.
Export or write back short notes answering the three core questions.

The Three Questions

Did it help? Did it make a decision clearer, easier, or more confident?

What was missing? Name the absent option, stale source, confusing term, broken route, or category gap.

Did you trust it? Where did trust increase or break: evidence links, scoring, language, privacy, or values control?

Optional fourth question: Was it calm? Did it feel finite and respectful, or cluttered and pushy?

Learning Receipt

For each tester or group, capture one receipt:

Path tested: individual choice, group decision, or instance stewardship.
Real decision: what they tried to choose, compare, move, fund, avoid, or build.
Did it help? What became clearer, easier, or more confident.
What was missing? The absent option, stale source, confusing term, broken route, or missing contribution path.
Did they trust it? Where trust increased or broke: evidence links, scoring, language, privacy, values control, exportability, or dissent.
Next patch: the smallest change that would help the next person.

The receipt should be useful even if the tester never creates an account and never shares their private values file.

What Counts As Earned Backlog

Turn feedback into work only when it is specific:

A missing mainstream option or values niche.
A stale or weak source.
A phrase that made the scoring feel more certain than the evidence allows.
A broken route, unclear first step, or failed share/preview path.
A repeated request from multiple testers.

Do not convert vague praise, vague dislike, or imagined scale into roadmap. The earned backlog must come from use.

Preview Email

Subject: Private preview: Values Commons

Hello,

I am sharing a private preview of Values Commons, a no-account, no-tracking way to choose by your values using sourced evidence.

Could you try it for one real decision and send back three notes?

Did it help?
What was missing?
Did you trust it?

Suggested path: start at the home page, take the two-minute tour, then choose one track: try one category such as banking, AI assistants, clothing, or learning resources; walk a group through Assembly and Slate; or use the Adoption Kit to scope a possible new domain. The preview is intentionally private and noindexed.

Thank you for testing the honest thing, not a polished pitch.

Stop Conditions

Pause broad sharing if any of these show up:

A tester reasonably thinks the app is sponsored or pay-to-rank.
A health, finance, or safety-adjacent claim reads as advice rather than a sourced comparison.
The preview package exposes source folders, raw internal docs, or search-indexable private pages.
Multiple testers cannot find the evidence trail.
A group cannot see dissent or a steward cannot find the Adoption Kit and contribution route.
A major first-use path fails on mobile or hosted preview.

Fix the trust break before inviting the next circle.