Mindforge
Six timed challenges, one IQ score — and an honest account of where it came from
Your saved progress: 0 XP across 0 games · view or move it to another device
About Mindforge
Mindforge is a brain test built as a game. Six sections, run back to back in one sitting: abstract pattern logic, number sequences, spatial rotation and paper folding, formal deduction, working memory, and raw processing speed. At the end you get a Brain Score out of 1000, a rank, and a breakdown showing which part of your thinking actually carried the run.
Four of the six sections are adaptive. Every item carries a difficulty from 1 to 10, and the next one is picked at the level you are currently holding — get it right and the test pushes; get it wrong and it backs off until it finds your edge. That is why the number of correct answers barely matters here. Seven right at level 3 is a weaker result than four right at level 8, and the game scores it that way.
Difficulty is not only in the puzzle — it is in the wrong answers. An easy item offers options that are plainly off. A hard one offers options that are each a single minimal step from the right one: a count off by exactly one, a shape swapped for one that looks like it, a piece with one cell moved one square, a mirror image rather than a rotation. At level 6 and above every wrong option in a pattern grid is one step away, so the rule has to be applied rather than eyeballed. The clock scales with the item too, so a hard one can be worked out on paper instead of rushed.
The items are the ones real cognitive tests use, rebuilt from scratch and generated fresh every time — thirteen kinds of them, rotating so that no two runs feel alike. Progressive matrices where up to four attributes change at once — and from the middle of the scale a rule may run down the columns rather than across the rows, so part of the work is spotting which way to look. Logic grids, the hardest item here, where the third pattern in each row is the first two combined by an operation you have to find. Knights-and-knaves puzzles where one group always lies and the only way through is to test every assignment. Odd-one-out figures where exactly one breaks the rule, checked at generation so no second answer is defensible. Visual analogies: the first pair changes by some rule, apply it to the third. Number series built from arithmetic, geometric, interleaved and recursive rules; balance scales where you have to work out what each shape weighs; and letter codes where one rule turns a word into its cipher. Mental rotation, where the tempting wrong answer is always a mirror image rather than a rotation. Paper folding, punched through and unfolded. Shape fitting, where a piece has been cut out and you have to find the one that fills the gap. Syllogisms and conditional chains written with nonsense names, so nothing you already know can substitute for the reasoning. A memory span that grows until you drop it and then flips to reverse order. And sixty seconds of symbol-to-digit coding, which is the one section where nothing is hard and everything is speed.
There is a Live Race too: up to ten players in one room, the same twelve puzzles, one shared clock, and a board that moves between every item. It ends on a scoreboard ranked by score and then by who thought faster, so being right is worth more than being quick but being quick still settles a tie. Empty seats are filled with stand-ins and they are labelled as stand-ins — a room that looks busier than it is would be a lie about who you are playing against. A race is reported as a sprint score rather than a Brain Score, because twelve fixed items across four skills is a rougher measurement than a full adaptive run and the two should not be quietly presented as the same number.
Nothing is memorisable, because nothing repeats: the puzzles are generated at the moment you see them. The one exception is the Daily Gauntlet, which is seeded from the date so that everyone playing it today gets the same items — the only way a score comparison between two people is actually fair.
The scoring is stated openly rather than implied. The game will tell you exactly how the number is built, and it will tell you plainly that the percentile sits on a reference curve rather than on a normed sample of human beings. It is not an IQ test, it does not claim to be one, and a browser tab is not where anyone gets diagnosed with anything.
How to play Mindforge
Pick a mode. Full Run is six sections of seven items, about eighteen minutes. Quick Run cuts it to four items a section. Daily Gauntlet is the full run on today's shared puzzle set.
Each section opens with a card explaining what it measures and how to answer it. Read it — the sections are genuinely different and the instructions matter.
Answer by tapping an option or pressing its number key. Every item has a clock, and it scales with the difficulty of that item — about 40 seconds on an easy one, up to 90 on the hardest — so a hard one can be worked out properly rather than rushed. Running the clock out counts as wrong, so commit rather than stall.
You will not be told whether an answer was right. You get your accuracy and your level at the end of each section instead.
Finish all six for your IQ score, the six-spoke profile, and a read of where your run was strong and where it was not.
Review on the results screen brings every item back with the correct answer marked, so you can see exactly where you went wrong. It runs after scoring, so looking changes nothing.
Open Rankings at any time for your own runs listed by Brain Score and how long each took, your last race, and the site-wide board.
About this game
More puzzle games
Recently added
Mindforge rankings
Ranked by best Brain Score. Scores are reported by the browser that played the run, like every board on this site — it is a bit of fun rather than a competitive ladder. Today resets at midnight IST, the week resets on Monday.
Loading…
Mindforge — frequently asked questions
What do I get at the end?
An IQ-scale score on the familiar mean-100 scale, the Brain Score it was derived from, a percentile, a rank, and a six-spoke profile showing which part of your thinking carried the run. Mindforge is built from the item types real cognitive tests use — progressive matrices, number series, mental rotation, paper folding, syllogisms, digit span, symbol coding — so the result is a real measurement of how you did on those tasks today, presented on the scale people actually compare.
What is a Brain Score out of?
One thousand. It is a weighted blend of your six section levels: patterns 22%, numbers 18%, spatial 18%, deduction 18%, memory 14% and speed 10%. Fluid reasoning carries the most weight because it is the part that generalises furthest beyond the test itself.
What does the percentile mean?
It places your score on a stated reference curve — mean 520, spread 150 — not on a normed sample of real people. The game says that in full on the results screen rather than dressing a modelled number up as a population rank.
Why does it not tell me if I got an answer right?
Because immediate feedback turns a measurement into a practice session, and the items that follow are chosen from how you are doing. You get your accuracy and your level at the end of every section instead.
Can I memorise the puzzles? How many are there?
There is no question bank to memorise — every puzzle is generated at the moment you see it. The pools were measured rather than guessed: the smallest runs to roughly 1,500 distinct puzzles, most run past 10,000, and none of them was exhausted after 30,000 attempts. The game also remembers the last 500 puzzles it has shown you and will not repeat one, so two runs in a row are genuinely different rather than reshuffled. The Daily Gauntlet is the deliberate exception: it is seeded from the date, so everyone who plays it on the same day gets identical items and the comparison is fair.
Is the IQ figure a real IQ?
It is your result placed on the scale people recognise — mean 100, spread 15 — and nothing more. It comes from the same measurement as the Brain Score shown beside it, positioned on a stated reference curve rather than on a normed sample of real people. A clinically administered assessment takes an hour or more, is given in person and is interpreted by somebody qualified to do it. This will put you in roughly the right place; it will not settle an argument, and it diagnoses nothing.
Can I play it against other people?
Yes. Live Race puts up to ten players in one room on the same twelve puzzles with a shared clock, a board that updates between every item, and a final scoreboard ranked by score and then by who thought faster. Empty seats are filled by stand-ins that are labelled as such — nothing in the room is passed off as a person. A race takes about nine minutes.
Is the race score the same as the Brain Score?
No, and the game keeps them apart. A race is twelve fixed items covering four skills, so it is reported as a sprint score: same 0-1000 scale, same estimator, but a rougher measurement, and no memory or speed section in it. The full solo run is the better measure of the two.
How long does a run take?
About eighteen minutes for the full run, or eight for the quick run. The clock on each item scales with how hard that item is — roughly 40 seconds on an easy one up to 90 on the hardest — so working something out on paper is possible rather than a race. The quick run uses four items a section instead of seven, so its score is noisier: good for a fast look, weaker as a measure.
Does it work on a phone?
Yes. Everything is tap-driven and the layout is built for narrow screens. A larger screen is a little kinder on the pattern section, where eight answer tiles have to fit side by side.
My score moved a lot between runs. Is it broken?
No — that is what a sixteen-minute measurement looks like. Sleep, mood, screen size and plain luck all move it. Two or three runs on different days tell you far more than any single number does.
Tell us what to fix
Found a bug, or think something should play differently? This goes straight to the developer — no account, no email address required.