Krillion Infinite is a free daily word game about obscurity, played in a browser. Every day you get seven prompts — name an actor, name a spice — and everyone in the world gets the same seven. Naming the first thing that comes to mind barely scores. The deeper into the water your answer lives, the further you sink.

It takes about four minutes. There is no account, no payment, no app and no advertising, and the whole thing runs on one page. If you want the rules rather than the reasoning, how to play is the shorter read.

Why the scoring runs backwards

Most quiz games reward you for knowing the answer. Krillion assumes you know an answer and asks a harder question: is it yours, or is it the one that a thousand other people typed into the same box this morning? Cheddar and Taleggio are both cheeses. One of them pays 10 points and the other pays 85, and the gap between those numbers is the entire design.

This is what makes a first dive feel like a failing grade. Seven correct answers can land you at 70 points out of 700, and nothing was wrong with any of them. That is the game working, not the game being unfair — and the fix is one habit, covered on the strategy page.

How a dive works

• Seven prompts, 25 seconds each. The clock starts the moment you see the prompt.

• A wrong guess costs you time, not the round — keep typing until the clock runs out.

• Answers are matched loosely, so a typo or a missing accent still lands.

• One dive a day. The next set unlocks at 00:00 UTC.

The six tiers

• 🌟 One in a Krillion — 100 · the single designated answer. Trench floor.

• 🦑 Deep Cut — 85 · true obscurity.

• 🐙 Rare — 60 · genuinely uncommon.

• 🐟 Schooler — 30 · solid, swims with the school.

• 🤡 Too Clever — 15 · the famously "obscure" pick everyone reaches for.

• 🫧 Plankton — 10 · the answer everyone blurts out.

Only a handful of answers per prompt sit at the bottom tier — the three or four a room full of people would shout in unison. Step off those and you are paid properly. Nearly four fifths of every bank is worth 85, because the hard part is not being obscure, it is thinking of anything at all in 25 seconds.

There is a seventh outcome that is not a tier: running the clock out. It scores zero rather than a penalty. Scoring has the shares each tier is cut at, the trap budget, and the band thresholds.

Depth

Every point sinks you 4 metres, so a dive is reported in metres rather than as a mark out of 700. The trench floor is 2,800m and needs a perfect dive: seven designated gems, seven times out of a hundred-odd answers each. Nobody gets a perfect dive. The five bands — Shallows, Schooler, Thermocline, Deep Cut, Trench — are pinned to real places in the water column you descend past on the way down.

Where the answers come from

Every prompt carries a hand-checked bank, sorted into how widely known each answer is. There are 73 prompts in the pack and about 1,550 accepted answers across them, averaging around 21 per prompt — the deepest runs to 50. Counting the alternative spellings and aliases each one accepts, the pack recognises close to 2,400 different things you can type.

That sorting is a judgement call, and judgement calls are wrong sometimes. A cheese you had to actually know should not be filed next to Cheddar, and an answer that half the country would give should not pay 85. If a good answer was rejected, or a common one is scoring as rare, tell us — rejected words are logged precisely so the bank can be repaired from what players actually type, and that mail is the most useful thing that arrives.

The percentile

The "deeper than X% of divers" line on the results screen is currently a model, not a measurement: it is estimated from your score, because there is not yet a real distribution of finished dives to compare you against. It will be replaced by real numbers once there are enough of them.