Adaptive Digital SAT mock exams with real module timing

Each module has its own countdown, and it is held server-side. Closing the tab does not pause it, reopening does not reset it, and a dropped connection does not cost you time you had already used — your answers save as you go and the remaining time is computed from when the module opened, not from what your browser thinks.

That matters for a reason beyond cheating. A practice test whose clock you can stop is not measuring pacing, and pacing is where most points are lost in the middle of the score range. The pressure is the measurement.

Once a module closes it stays closed. You cannot go back to Reading and Writing module 1 from module 2, because on the real test you cannot, and a mock that lets you is quietly teaching a strategy that will fail on the day.

The runner is not a question and four radio buttons. It carries the working surfaces the real test gives you, because a student who has practised without them is learning to sit a different exam.

The highlighter works over the passage and over the question itself, and your marks are saved per question, so flagging something and coming back to it does not hand you back a clean passage to re-read from scratch. The line reader dims the page and leaves one band of it lit, which is what people who track lines with a finger have been doing on paper for decades.

There is a scratchpad, and a note you can leave on a single question. Both persist through the module, so a refresh does not destroy your working.

The real Digital SAT embeds Desmos, so a calculator that works like Desmos is what a student should practise on. We could not ship Desmos itself: its API gives no free commercial use, and the grant is revocable, so a key rotation could take the calculator away from every student mid-exam at once.

So we built our own, to do what the SAT's calculator does. It graphs lines, curves, circles and inequalities as you type; a tap on a graph shows its intercepts, its maximum or minimum and where it crosses the others, which is the fastest honest way to solve a system on the test. It has sliders, tables and regressions, degrees and radians, and it shows a fraction when an answer is one.

It is not Desmos, and we say so: some buttons are in different places, and expressions are typed as text rather than into a maths editor. Twenty minutes in Bluebook's official practice before test day closes that gap.

There are three full-length forms. Each is 147 questions across six panels — a first and a second module for each section, and the second module exists twice, once easier and once harder, because a student routed up and a student routed down must both have somewhere to go. Two of the three are free.

A form is published only when every one of those six panels is full. A published form with an empty panel is an exam that starts, runs, and then fails to score, which is worse for the person sitting it than a form that was not there. For most of this product's life that rule was the reason the mock list was empty, and saying so was more useful than a launch-timing excuse.

The fourth form is being written against the same blueprint. It will appear the day its last panel closes, and the count on this page will change in the same commit.