gym for your mind.

Suggestions come from what you have already done: skills you have barely played, skills with lower ratings, games you have not tried, and words due for review. What you played recently keeps the suggestions varied. If you only read Standard passages, we compare Language against their ceiling of 60, so that lower cap alone does not make it your weakest area.

Until you finish today’s lock-in, “Recommended” points to its next workout.

If your latest test shows an area to work on, Home can also offer “Focused practice,” marked “Based on test.” Once you have done a follow-up, we stop recommending it.

Try the controls, watch what changes, then get every tank to its target. The same controls work for all three targets. What they do stays hidden. Working that out, and remembering it, is the point.

Each mark is one unit. A control only works if there is enough liquid, and enough space, for its full action. Press A, B, C, or D for the controls and Z to undo. Undo takes you back one step. Reset takes you back to the start of the target. Presses you have used still count in a challenge.

Exploring is not scored. Each target has a live timer that leaves out pauses and the time between targets. Giving up ends a target with no credit. Solutions stay hidden until you have finished all three, so one target cannot give away the next. After that, you can step through a shortest sequence of presses for each one. Your round result is your average efficiency across the three targets: the shortest possible solution divided by the presses you used, as a percentage. Higher levels change how the system works, not just how fast it goes.

Send each piece through the gate that matches. The banner tells you what to match: color or shape. The rule changes while you play. A round is 60 seconds of active play, and you can pause it.

Points and streaks are just for that round. Your rating looks at accuracy, difficulty, and how many decisions you made. Waiting around and answering once will not earn a high rating.

Every arrow moves both pieces. A wall only stops the piece next to it. Use that difference to get both pieces onto their flags at the same time. There are three puzzles and no time limit. A live timer tracks each puzzle, and your results and activity history show active solving time only. Pauses, time with the tab hidden, time spent viewing solutions, and time between puzzles are left out. Time does not affect your score.

Efficiency compares the shortest solution with the moves you made, averaged across the three puzzles. Undo counts as a move, and reset does not take back moves you have already made. You can see the solution once you solve a puzzle or give up. Giving up ends that attempt with no credit and stops its timer, and you cannot go back to it. There are no hints during play.

Answer six questions using only what you are told. There are five levels, from Beginner to Expert. Questions cover arguments, assumptions, deductions, and puzzles where the constraints change.

We wrote these questions ourselves. Some follow logical-reasoning formats like the LSAT, but they are not official LSAT material, and the levels are not calibrated to LSAT difficulty. Questions you have not seen come first. Older ones come back once you have used up the pool.

Confidence is optional, in steps of 5%. Nothing is recorded until you pick a value. It locks when you answer and never affects your rating. You can keep editing notes after you answer. Both are saved with a finished round.

Each round has six mental math and reasoning puzzles, across six levels. They lean on shortcuts, patterns, everyday calculations, and putting clues together. Try them in your head, but writing things down is fine.

Get five or more right and we suggest moving up a level. Get two or fewer and we suggest moving down. Your rating uses accuracy and difficulty, not speed.

Watch the square and listen to the letter. At 2-back, compare each one with two turns ago. Press A or Position if the square matches, L or Letter if the letter matches, or both. If neither matches, do nothing. You can choose anything from 1-back to 6-back.

There are 30 turns. Your browser speaks the letters. If it cannot, they show above the grid. Matches and non-matches count equally, and so do position and letter. That way, pressing nothing cannot earn a strong score. Balanced accuracy of 50% or less earns no rating credit.

Meet 3, 4, 6, 8, or 10 people, with eight seconds each. Pass a target game, recall their names and jobs, pass a coin-catching game, then recall them again. You need to pass both games in between, and you can retry them.

The 46 portraits are generated, fictional people, and we chose their names. Each face keeps its name, but jobs are shuffled every round. Capital letters and spaces do not matter in name answers. Corrections, including jobs, appear at the end.

Both recalls are saved, but your rating comes from the second recall and how many people you met. Scores from the games in between do not count. Leaving the page or hiding the tab ends a round you have not finished.

Read a passage, then answer a question about it. Nonfiction links to its sources. Fantasy is original fiction. New passages come first. Summaries are optional and not graded. Only your first answer to a passage counts toward your rating. Repeating it is just practice.

Tap a highlighted word to save its definition and schedule a review. It comes back in a new sentence the next day. Each time you remember it, the gap grows to 3, 7, 14, then 30 days. If you forget it, it comes back tomorrow. Your saved words are in Progress. Remembering words does not add rating points.

There are two levels. Standard passages are short and direct. Advanced passages are longer and denser, and their questions ask you to connect ideas across the passage rather than recall one detail.

Logic and Numbers time each answer, including time away from the tab. The timer pauses while you are writing a Logic note or setting your confidence, and stops when you submit or skip. Reading the explanation and editing after you answer are not timed. Round time is the total of the question times. Finished Logic and Numbers rounds save the questions, your answers, the correct answers, and the explanations, so you can review them in Progress.

Reading times how long you spend reading and answering, and pauses while the tab is hidden or you are writing the optional summary. N-back records its round time and a review of every turn. Face & Name times the whole workout, including introductions and retries of the games in between, plus each recall on its own. You can reopen its name and job corrections from Progress. Rule Rush saves a review of your decisions. Other game ratings do not reward speed. Rounds you do not finish are not saved as completed activity.

At each level, we average a quality score between 0 and 1 across your recent rounds. Then:

Rating = difficulty ceiling × quality² × evidence factor

The evidence factor is 0.7 + 0.3 × min(1, samples ÷ minimum), using the minimums above. For example, three perfect Level 1 Logic rounds give a game rating of 20.

Logic adjusts accuracy for the number of answer choices. Reading adjusts its average accuracy for a 25% chance of guessing right, and n-back and Rule Rush adjust for 50%. Adjusted values are kept between 0 and 1.

Numbers and Face & Name use accuracy as it is. Side Effects and Twin Run use solution efficiency, and giving up scores zero for that puzzle or target. Older Twin Run rounds that used hints keep their original cap of 0.5. Rule Rush multiplies adjusted accuracy by min(1, decisions ÷ (30 + 5 × (level − 1))). A Rule Rush round only counts if it has at least 55 seconds of active play.

We keep your highest estimate across levels before rounding, then round the game rating you see. Skill ratings average the longer-history game estimates after rounding them.

A round can add to your record without beating your strongest level. Rounding can also hide a small change. Your activity history keeps each round result and shows your game and skill ratings before and after.

Earlier results still count, except older game formats, incomplete records, and reading marked as assisted. The test does not count toward game or skill ratings either.