Questions

What is Caste

A social network that sorts its users by assessed ability and keeps them within their assigned stratum. You take a short assessment when you join. You are placed in one of four castes. You post and read within your own caste only. Your placement is recalculated weekly.

Why

Every social platform ranks its users. Reach, visibility, and who sees what are all decided by systems that are never disclosed and never named. The ranking exists whether or not you are told about it. Caste makes the ranking explicit and lets you see the mechanism you are already subject to elsewhere.

Connection also works better between peers. Most platforms optimize for reaction, not for finding the people you actually want to talk to. Stratification is one answer to that.

The assessment — what it is

Two parts, thirteen questions, individually timed, one attempt.

Part I · Shape Rotator — six figural items with no language, drawn from the standard non-verbal reasoning families:

• Matrices — a 3x3 grid follows a hidden rule; you supply the missing cell. This is the format of Raven's Progressive Matrices, developed in 1936 and still the most widely used non-verbal measure of abstract reasoning. Matrix reasoning is among the most g-loaded single tasks in psychometrics.
• Series — a sequence of figures follows a rule; you supply the next. The completion format used in the Cattell Culture Fair Intelligence Test, designed in the 1940s specifically to reduce the effect of language and schooling on scores.
• Classification — five figures share a property; one does not. Also from the Cattell battery.

Part II · Word Cell — seven items. Four are verbal — analogies, ordering, classification, and self-referential parsing. Three are a word span: a short sequence of words is shown one at a time, then reproduced in order — a span task in the tradition of digit span, which has appeared on the Wechsler scales since 1939 and measures working memory rather than reasoning. Verbal reasoning and working memory each measure something different from figural reasoning, and both are included so that the composite does not rest on a single ability.

Why these formats

Non-verbal reasoning tasks were built to be usable across languages and educational backgrounds. They are not neutral — no assessment is — but they depend less on vocabulary, schooling, and cultural reference than verbal or knowledge-based tests do. Every figural item on Caste is procedurally generated and human-reviewed, so there is no answer key to circulate. No two assessments are identical.

Is this an IQ test

It uses formats drawn from the same literature. It is not a clinically administered instrument, it is not proctored, it is not normed against a standardized population, and it produces no IQ score. It is a thirteen-item assessment taken in three minutes on a website. Treat it accordingly.

How scoring works

Items are weighted by difficulty. Response speed contributes a small amount and can only break ties, never outrank accuracy. Section scores are combined into a composite weighted by each section's established relationship to general ability. Your placement is your composite's position relative to everyone else assessed under the same test version. Scores are never displayed.

What else counts

Placement is not the assessment alone. Account signals from the platform you sign in with, activity on Caste, and the equipment you arrive on all contribute. Weights are not published.

How to advance

Placement is recalculated weekly and can move in either direction. Advancement methods differ by caste and are shown to you if they are available to you. Some are effort. Some are recruitment. Some are purchase. Not all castes have one.

Can I retake the assessment

No. One attempt, ever.

How reclassification works

Placement is not a verdict returned once. Every week the system recomputes your standing and re-sorts the whole population, so your caste is a live position in a moving distribution rather than a permanent label. Two kinds of signal feed that computation.

The fixed term is your assessment — a one-time measure of reasoning and working memory. It sets a baseline and never changes; it anchors where you begin.

The moving term is what you do afterward, and it is what actually carries you between strata. It scores demonstrated judgment over time rather than a single sitting. Two forecasting markets hold most of that weight:

• Stocks. Your equity calls are scored against the market itself — the S&P 500 (SPY) over the same window. Beating the benchmark is evidence of signal; trailing it is evidence of noise. The index is an adversarial baseline that most professional managers fail to beat over time, so consistent outperformance is a costly and therefore credible signal. Luck regresses to the mean; skill persists. Because you are scored weekly across many calls, no single lucky pick can lift you and no single bad week can sink you — what is measured is your calibration over the long run.
• Netflix. You commit a prediction of the Netflix weekly Top 10 before it publishes. This measures a different faculty entirely — not financial reasoning but attention: the capacity to model what a mass audience will actually watch, as distinct from what you think they should. Reading a crowd accurately means setting your own taste aside, and it correlates only loosely with the first market. A person can be sharp on equities and blind to culture, or the reverse; measuring both resists a one-dimensional read.

Both are proper forecasts: you commit in advance, reality settles the outcome, and you are scored on the gap between them. There is no credit for hedging and none for confidence that accuracy does not back. Account signals, your activity within your caste, and the equipment you arrive on account for the remainder. Weights are not published.

When the recomputation moves you across a boundary, the change is immediate. Your feed, your search, and your connections re-scope to the new stratum, and relationships that cross the new boundary become inactive. You are not notified of the reason.

Privacy

See the privacy policy. Assessment responses, timing telemetry, account signals, device and browser data, and direct messages are stored. Direct messages are not end-to-end encrypted and are readable by administrators when reported.

Moderation

Reports are reviewed. Accounts may be removed. There is no appeal process.

Who made this

A consortium of outcasted Brahma.

Contact

hi@caste.social

Sources

SOURCE: NETFLIX WEEKLY TOP 10. Chart posters are provided by TMDB. This product uses the TMDB API but is not endorsed or certified by TMDB.