04 · breakthetest
Break the Test
Practice the test you actually take.
A pattern-recognition trainer for the digital SAT, ACT, AP exams, and Common App essays. Five-minute rounds that name the traps quietly costing students points, on an item bank that does not run out.
The thinking
The test inside the test is the part that costs students points.
Who it's for
High school students preparing for the digital SAT or ACT, AP exams, or college application essays. Tutors and parents who want a practice loop that does not run out of new material.
Why it exists
Two problems made this tool necessary. First, students keep losing points to the same hidden structures inside test questions: the answer that overclaims, the wrong answer that paraphrases background text, the prompt verb they misread. They do not lack effort. They lack a way to see the pattern. Second, the official supply of practice tests is rationed. The College Board publishes a handful of SAT versions, and serious students burn through them in weeks. Everything after that is recycled material or low-quality knockoffs.
How it helps
Break the Test trains one narrow pattern family at a time. Five-minute rounds on a phone. After each answer the tool names what the wrong choices were doing, so the pattern becomes visible the next time you see it. The item bank is large and continuously refreshed, so rounds keep producing material the student has not seen before.
About the intelligence
The intelligence is in the feedback. After every answer the tool names what each wrong choice was doing, so the pattern becomes nameable instead of a vague feeling. The method is in the background. What matters is whether the next question becomes easier.
What it trains
Twelve modules. One pattern family each.
Each module trains one narrow skill: a kind of trap, a kind of prompt, a kind of structural read. Five-minute rounds. Feedback that names what the wrong choices were doing.
20,130 items across the twelve, every one of them checked by more than one independent method before it shipped. Counts as of 3 October 2026, and they only go up.
Digital SAT
Trap answers, sentence boundaries, choosing a representation instead of grinding the algebra, and reading a figure for what it was actually asked.
- Math Strategy5,978
- Reading4,102
- Grammar2,411
- Data & Figures2,175
ACT
Four modules, and full-length tests scored the way the exam scores them, 1 to 36.
- ACT Reading1,182
- ACT English909
- ACT Science577
- ACT Math487
AP
Reading the rubric verb before the prompt on free-response, plus a rules-fluency trainer that behaves the same way every time.
- AP Essays819
- AP Calculus350
- AP Precalculus105
Application essays
What a supplement is actually asking, worked out before a word gets written.
- Application Essays1,035
Beyond the rounds
The official supply runs out. This one does not.
A student can start a full, timed test with a real section break between reading and math, and the clock stops while they are away from it. As of 3 October 2026: 64 SAT forms, 9 ACT forms, 8 PSAT forms, and 2 AP Calculus AB sections, each one a complete section a student can sit down and take.
Leave a test unfinished and it reports a range rather than a score it did not earn. Measured against finished tests, that range carries roughly 86 points of error at ten answers, 67 at thirty, and 43 at sixty. It is framed as where this test was heading, never as what test day will hold, because those are different claims and only one of them is honest.
Every test ends with five questions in the skill the student missed most, and not one of them comes from the test they just took. A student who does not want to open with a full test can open with fifteen questions instead.
Closing note
Every item earns its place.
Every question is solved from scratch and independently checked before it ships. Mathematics is verified symbolically. Where the checks disagree, the question does not go out.
On 2 October 2026 a student told us a question was confusing. Its setup asked for the largest integer satisfying a condition, then the prompt asked only for the value, so a student who answered the prompt as written gave the boundary value and was marked wrong. The student was right.
The answer key had been correct the whole time, which is why every answer-checking gate passed it. The check that should have caught the wording had flagged it months earlier and then counted it into a baseline, because a measure that only asks whether there are more problems than before can never ask whether one of them is wrong. It is fixed, that shape is blocked now with no baseline to hide in, and the authoring guide was rewritten. We would rather publish the miss than the claim.