One choice at a time. Many ways to connect.

Start with the five-question WhatsApp Quick Read.

Pick five everyday answers and create one private link. The invited person predicts yours, adds their own, and reveals Knowledge plus Match. Their result can turn the five answers they already made into a one-tap return challenge. Sharing is always explicit; the completed result is never locked behind a post.

Make the 24-question Know Me quiz.

Answer 24 everyday choices once and create one private link. Each friend predicts your answers with one tap per question. Their reveal shows correct/24, six category reads, strongest reads, plot-twist zones, and every actual answer.

This mode is prediction-only for the friend, so there is no second self-answer phase and no Match score. After the reveal, “Flip 8 back” takes the best plot-twist questions straight into an eight-tap return challenge—no pack or question selection screen.

Or build a custom private deck.

Choose 5–30 items from the reviewed 100-question bank, answer them once, and share one fragment link. Friends predict locally and reveal one exact score. The deck and answer key stay in the URL fragment rather than a ViewsMeet database; link holders can technically inspect it, so it is private-by-sharing rather than encrypted.

Let a whole group answer your chosen questions.

Group Picks freezes 5–20 questions from the reviewed bank into one private link. Everyone answers independently. Question-level splits appear only after three retained responses; individual answer sheets, names, relationship labels, and cross-question profiles are never shown. Rooms accept at most 50 responses and expire after 14 days.

Play first. Sign in only to keep it.

ViewsMeet has an optional passkey account for the quiz, trait, and Group Picks links you explicitly save. It can reopen invitations and creator summaries on another device where your passkey is available. There is no email, password, public handle, or support recovery. Private deck fragments stay only in the browser that saved them and are never synchronized.

Map five traits, then let someone predict your self-ratings.

Complete the exact public-domain Mini-IPIP 20 and receive five deterministic raw means. You can then create a seven-day private link for a friend or software agent. They predict your 1–5 rating for the same exact first-person statements.

Both 20-item vectors are discarded immediately. Challenge storage keeps only five keyed sums per role; the reveal shows five paired means and signed gaps. It contains no item reveal, overall accuracy, rank, type, percentile, diagnosis, or compatibility score. The prediction layer is an unvalidated perception game—not a validated observer assessment.

1. Pick what you would choose.

Today begins immediately with five curated either/or questions. There is no account and no “correct” self answer.

Or take the 21-choice full profile.

One tap per trade-off builds seven transparent “this round” axes. Share a private profile challenge and a friend can complete the same round to reveal where your patterns align and diverge.

2. Predict the named Lab panel.

After each personal choice, guess which side the clearly labeled target panel will take. The panel configuration is visible; the label is not an identity-verification claim.

3. Make one private friend link.

After the result, create a challenge with an optional alias. A friend first predicts your five answers and then records their own. Anyone with the URL can participate until it expires, so share it intentionally.

4. Read Knowledge and Match separately.

A completed per-question comparison can show the creator’s issued choices, but never another respondent’s sheet. Neither score labels the friendship. Safe reactions keep the response lightweight: no chat, insults, or public profiles.

5. Let a private link grow carefully.

More friends can answer the same link. Exact privacy-safe group tallies appear at five total people, including the creator. Below five, the view remains visibly forming. Individual respondent sheets and least-compatible rankings remain out of bounds.

6. Opt into categories—not a public social graph.

The private creator board can show optional relationship and person/software labels for one challenge. Public Atlas permission is separate and off by default on each side: a participant cannot publish the creator’s private actor category. Once daily, only category cells with at least 20 direct edge occurrences from 20 independent challenge lineages can open. Names, links, answers, scores, locations, and person-to-person nodes never enter the public document.

7. Give an agent the same bounded choices.

REST, MCP, A2A, OpenAPI, and the published Agent Skill expose strict enum-only operations. An agent can read public packs and frozen Atlas cells. It can answer a private Know Me or personality-perception challenge only after an operator supplies that exact capability. Transport proves neither model identity nor autonomy, and no tool automatically posts a returned link.