This isn't a personality test.
It's a cognitive profile — built from what you do, not what you say about yourself.
20 to 50 scenario questions.
No right answers. No wrong answers.
Just decisions — and what those decisions reveal about your default way of processing complexity.
You'll get your top 3 thinking styles, ranked. Plus insights on how they interact.
Not people who share your job title. People who share your cognitive fingerprint.
This test doesn't label you. It gives you language for what was already true.
If something feels off, take it again — and this time, pick the honest answer, not the impressive one.
This test uses behavior-based scoring, not self-report. Every question is a real scenario. Your answers carry weighted signals across 10 thinking styles.
The algorithm is adaptive: once a style is confidently identified, the test pivots to probe your remaining styles. Sessions naturally run 20–50 questions.
No personal data collected. Sessions are anonymous via a local token. No cookies. No consent banner required.
Results show your top 3 thinking styles, ranked by strength, with a breakdown of how they interact.
THINK! — A FREEBIE BY TGS — THEGREATSTAG.COM
A behavior-based cognitive profiling tool. Not a personality test. Not a performance test.
It doesn't ask how you see yourself. It puts you in situations and watches how you decide.
You get 20 to 60 scenario questions drawn from a 100-question bank. Your answers are scored across 10 thinking styles. At the end: your top 3 styles, the cognitive signature that comes with them, and 3 characters from history and fiction who share your profile.
No account. No personal data collected. No AI in the runtime. Just your decisions revealing how your brain works.
Think[!] profiles across 10 styles. Not personality types. Not how smart you are. How your brain defaults to processing complexity.
Each style has a name, a core behavior pattern, a cognitive signature, strengths, blind spots, and a “pairs well with” mapping. All documented publicly.
These 10 styles didn't come from nowhere. The academic and conceptual grounding: cognitive science, decision theory, existing frameworks studied, why we kept what we kept, and why we left out what most tools include.
100 questions. Each one is a scenario, a trade-off, or a prioritization challenge. Every question has weighted answer options. Each option carries a distribution of points across multiple styles. You rarely score only one style per answer — most responses reflect a blend.
| Status | Draw weight | Meaning |
|---|---|---|
testing | 0.40 | New, needs calibration data fast |
active | 0.20 | Established, reliable signal |
low_signal | 0.10 | Weak discriminator, being phased out |
retired | 0.00 | Never drawn again, data archived |
No two sessions are identical. Questions are drawn probabilistically each time.
135 characters. Real people and fictional. Each one mapped to a primary and secondary thinking style.
At the end of your test, your full top-5 style profile is scored against every character in the database. Primary style match: 3 points. Strong secondary match (S+): 2 points. Secondary match (S): 1 point. The 3 highest-scoring characters appear in your result. Two people with the same dominant style can get different characters.
Character images are fetched live from the Wikipedia API. No images are stored.
The test doesn't run for a fixed number of questions. It runs until your top 3 styles are “locked.” A style locks when all 3 conditions are true:
1. You've picked it at least 6 times
2. You've picked it significantly more than the next unlocked style (1.35× for slot 1, 1.2× for slots 2 and 3)
3. OR condition: you've picked it 10 or more times with a gap of 3 or more, regardless of ratio
Minimum 12 questions before any style can lock. Hard cap at 65. If the engine hasn't locked all 3 slots by then, it outputs whatever is current. Ambiguity is a valid result.
No AI in the runtime. The scoring algorithm is pure client-side JavaScript. Content is fetched from a database at page load via a Cloudflare Worker proxy. After that, everything runs in your browser.
What's collected: a temporary anonymous session token, your answers and their scored weights, your final style scores, and interaction signals for calibration quality scoring only.
What's not collected: your name, email, IP, or any personal identifier. No cookies. Nothing that could link a session to a real person.
Anonymous session tokens exist to prevent duplicate submissions from inflating the global style counter. They are not tracked across sessions or devices.
The question bank evolves. Bad questions get retired. New ones get promoted. Users who take the test more than once and get different dominant results see a prompt: “Which result feels most accurate?” Those signals accumulate. When 200+ meaningful confirmations exist, a calibration run analyzes which questions predicted the confirmed styles — and which didn't. Draw weights adjust. The cycle resets.
No personal data is used in calibration. All signals are aggregated and anonymous.
| Version | Output | Hard cap | Notable changes |
|---|---|---|---|
| Beta v1 | Top 5 | 50 Q | First working engine. All content hardcoded in JS. No draw weights. |
| Beta v5 | Top 5 | 60 Q | Absolute score + gap locking per slot. Still fully hardcoded. |
| Beta v8 | Top 5 | 50 Q | Supabase + Cloudflare Worker. Session writing, troll detection, quality scoring. Draw weights introduced. |
| v1.0 (live) | Top 3 | 65 Q | Output cut to top 3. All content migrated to Supabase. Engine fully content-free. Multilingual architecture. |