Free concept playground · no account

What Is It Actually Grading

Answer interview questions, then see what was really scored.

What Is It Actually Grading

AvailableFree

Paste an interview answer. You’ll assume it’s graded on how polished and thorough it reads — the screener ignores prose entirely and scores four hidden structural signals. See which ones it actually read, and which it never saw.

50/ 100 hidden scoreMixed read — half the rubric is missing.
2 of 4 signals landed
  • StructureOrdered reasoning — first / then / finally, or 3+ sentences.
  • SpecificityA concrete anchor — a number, a %, or a named thing.
  • OwnershipYou did it — "I" / "my", not "we" / "the team".
  • OutcomeA result, not activity — shipped, reduced, increased, %.
What it was actually gradingthe rubric, not the polish

Most candidates think length and eloquence win — a long, polished paragraph feels like a strong answer. The screener never scores prose quality; it checks these four structural signals and nothing else.

  • Never saw: StructureOrdered reasoning — first / then / finally, or 3+ sentences.
  • Never saw: SpecificityA concrete anchor — a number, a %, or a named thing.
  • Read: OwnershipYou did it — "I" / "my", not "we" / "the team".
  • Read: OutcomeA result, not activity — shipped, reduced, increased, %.

You lost points on Structure, Specificity — not because the answer read badly, but because those signals were never on the page.

Share on WhatsApp
Was this playground useful?

What Is It Actually Grading

Paste an interview answer and you’ll assume it’s scored on how polished, thorough, and eloquent it reads. It isn’t. An interviewer — or an AI screener — ignores your prose quality and checks four structural signals: ordered reasoning, a concrete anchor (a number, a %, a named thing), personal ownership (“I” / “my”, not “we”), and a stated outcome. This playground scores those four, hands you a 0–100 hidden score, and shows the exact signals it read versus the ones it never saw. The reveal is the gap between what candidates think is graded (length and polish) and what actually moved the needle. Runs entirely in your browser (0 uploads, works offline).

How to use it

  1. Paste your answer into the box, or load a preset — Strong answer, Eloquent but vague, or Wall of text.
  2. Read the hidden score and the four signal chips: a green check means the screener read that signal, a cross means it was invisible.
  3. Open the X-ray panel to see what you thought was graded versus the four signals that actually were — and which ones cost you points.
  4. Edit your answer and watch the score move. Adding one number or one “I shipped” often does more than a paragraph of polish.

What this clears up (the fundamentals)

  • Eloquence is not a signal — a long, warm, well-written paragraph can score near-zero. The rubric never reads prose quality; it reads structure, specifics, ownership, and outcome.
  • “We” erases you — ownership is scored on “I” / “my”. An answer full of “we collaborated” and “the team” leaves the interviewer unable to tell what you did.
  • A number beats an adjective — “much faster” is invisible; “40% faster” is a concrete anchor. One metric or one named thing flips the specificity signal on.
  • Activity is not outcome — “I worked really hard on it” describes effort; “I shipped it and error rates dropped” describes a result. The screener scores the second, not the first.

Where it’s used

A top-of-funnel career reliability check — the fastest way to feel why a technically fine answer still fails: judgment and signal, not correctness, are doing the grading. It’s a miatz build-lab concept playable: play it here, then learn to build the signal detectors and the judgment-vs-polish reveal yourself in the Interviewing on Judgment, Not Just Output lab.

FAQ

How is the hidden score calculated?

Four signals are each a yes/no on your text: structure (ordering words like first / then / finally, or 3+ sentences), specificity (a digit, a %, or a mid-sentence named thing), ownership (“I” / “my”), and outcome (result words like shipped / reduced / increased, or a %). Your score is round(signals-landed / 4 × 100), so 0, 25, 50, 75, or 100. It’s a deterministic rubric — no model call, no randomness.

Why did my long, detailed answer score so low?

Because length is not one of the four signals. A wall of text with no number, no “I”, and no stated result lands only structure — 25/100 — no matter how fluent it reads. That gap is the whole lesson: polish feels like the grade, but it isn’t on the rubric.

Is this the same rubric a real interviewer uses?

It’s a teaching model of the same idea — interviewers and AI screeners reward structured, specific, owned, outcome-led answers. Real panels weigh more and judge nuance, but these four signals are the ones candidates most reliably forget, so scoring them makes the hidden rubric visible.

What’s the fastest way to raise my score?

Turn one claim into a number, and one “we” into “I” with a result. “We improved things” → “I shipped a cache that cut p95 latency 40%” flips three signals at once.

Is anything uploaded?

No. Your answer is scored entirely in your browser — nothing is transmitted, stored, or logged. Turn off your Wi-Fi and it still works.

Limits

A teaching rubric with four fixed signals and simple text heuristics — it’s an opinionated read on answer shape, not a hiring decision. It can be fooled (drop in a stray number and specificity flips on), it doesn’t judge whether your reasoning is correct, and a real interviewer weighs relevance, depth, and honesty this can’t see. The signals it names, though — structure, specifics, ownership, outcome — are exactly the real practice.

Related

Part of the Demystify Playgrounds. Explore the rest from the Playgrounds home.

Bookmark this page (Ctrl+D, or ⌘D on Mac) — it works offline the next time you need it.

Ninety playgrounds. Zero setup.

Every concept here is playable free, no account — and inside the program you learn to rebuild the machinery yourself.