How we fit the SAT into a text message

The digital SAT is a screen-based, adaptive, three-hour exam with graphs, data tables, reading passages, and a built-in graphing calculator.

A text message is 160 characters.

Reconciling those two facts is most of the engineering behind TextPrep, and the interesting part isn't the technology. It's the judgment calls. Every one of them comes down to the same question: what can we drop without changing what the question actually tests?

Here's how each format works, and why.

The rule we work from

Fidelity to the question, not to the interface.

Nobody needs their phone to reproduce the Bluebook testing app. That's not what daily practice is for. What a student needs is to encounter a real SAT-style question, think about it the way the test wants them to think, and find out whether they were right or wrong, and why.

Which means we'll cheerfully give up the interface — the timer, the on-screen highlighter, the module structure — and we'll fight hard for anything that changes the reasoning. A graph you can't see changes the reasoning. A five-minute timer doesn't.

That principle sorts almost every decision below.

Multiple choice: the workhorse

Most of what we send is standard multiple choice. Question, four options, reply A, B, C, or D.

This is the overwhelming majority of what we send, and it survives the trip to SMS essentially intact. There's nothing lost in translation: the student reads the same words, weighs the same four options, and commits to one. That last part matters more than it sounds. Tapping a radio button and typing a letter feel different; typing feels like a decision.

The real constraint here is length. Your phone carrier (Verizon, AT&T, T-mobile, etc.) split anything past 160 characters into multiple segments, and a full Reading & Writing question with four answer choices routinely runs to two. We cap at three. That ceiling shapes which questions make it into the bank at all: an item whose stem sprawls past three segments either gets tightened or doesn't ship. It's a hard constraint, and it's made the question bank better. SAT questions are rarely improved by extra words.

It also shapes how the explanation reads. When you reply, the worked explanation comes back in the same thread, under the same ceiling, which means it has to say why the right answer is right and why the tempting wrong one is wrong without padding. Most of the editorial work on this product is that: cutting an explanation down until only the reasoning is left. That's the part students actually see every day.

Student-produced response: where text is actually the better medium

The SAT's math section includes questions with no answer choices at all. You work out the answer and enter it. College Board calls these student-produced response questions (SPR); most students know them as grid-ins.

These are the ones where SMS stops being a compromise and starts being an advantage.

On a multiple-choice question, a student who's unsure has four options to work backward from. Plug in choice B, see if it holds. That's a legitimate test-taking strategy and we're not against it, but it means multiple choice always tests a blend of "can you solve this" and "can you eliminate."

An SPR question texted to a phone strips that away completely. There's nothing to work backward from. You either produce the number or you don't, and you type it into the reply box the same way you'd type it into Bluebook. No format loss, and a cleaner read on whether the student actually knows the material.

Graphs and figures: MMS, not description

Some SAT math questions are built on a reference visual: a scatterplot, a bar chart, a data table, a geometric figure. There's no version of "describe the graph in words" that preserves what the question tests. Reading a chart is the skill. The moment you translate it into prose, you've replaced a data-interpretation question with a reading question and thrown away the point.

So when a question needs a figure, the figure goes out as MMS, a real image on the student's screen, the way it appears on the test, rather than a paragraph describing it. The system composes the message and attaches the asset the same way it assembles any other question.

Image delivery has more ways to fail than text, and the failure is quiet: a broken asset can still look like a successful send from our side. We carry text descriptions as an accessibility fallback, and we'd rather eat the cost and the complexity than send a student a question about a graph they can't see.

Long passages: one tap away

Reading & Writing questions built on longer passages are the hardest case, and it's the one we argued about most.

Three options, all imperfect:

Inline the passage. Simple, no second surface, works on any phone. But a 140-word passage means a long, expensive message, and re-reading it means scrolling back up through a thread. Reading comprehension requires actual re-reading, and a text thread is a hostile place to do it.

Send it as an image. MMS makes this technically possible. It's the wrong answer: a screen reader can't read text baked into a picture, and re-reading means pinch-zooming small type. Text is the native form for text.

Put it on a page and link to it. The passage lives at its own URL. The text message carries the question and the answer choices; the link carries the reading, on a page built to be read and to work with a screen reader.

We built the third. It's the only format that asks a student to leave the message thread, which is a real cost (a tap, a page load, a switch back), and we took it because the alternatives were an unreadable wall of text or a passage a blind student couldn't access.

What we deliberately don't do

Two things we could build and probably won't:

We don't adapt the schedule to performance. Every student gets one question a day on a fixed cadence, regardless of how yesterday went. There's a whole category of software that shortens intervals on questions you miss and lengthens them on questions you nail. We're not doing it. Partly that's a deliberate boundary we've drawn and intend to hold, and partly it's that a predictable daily rhythm is the thing that makes the habit stick. A schedule that speeds up when you struggle is a schedule that punishes you for a bad week.

Worth being precise about what that does and doesn't rule out, because the distinction is the whole design. We're planning to let students set their own difficulty: you'd tell us you want harder or easier questions, and we'd send them. That's a setting you control, not an inference we draw. The system would never move it for you, in either direction, based on how you're answering. You choose; we don't change it behind your back. (That one's on the roadmap, not in your phone yet.)

We don't simulate the full test. No timing, no adaptive modules, no three-hour sitting. Those are essential and Bluebook's official practice tests do them properly and for free. Trying to reproduce them over SMS would produce a worse version of something that already exists. We fit in between those practice tests.

Balancing the two sections

One last decision that isn't about format at all.

The digital SAT weights Math and Reading & Writing equally in the composite score, so we send them equally, a 1:1 split across a student's run. If student feedback supports it, we'll build in the ability to weight your questions the way you want.

One real SAT-style question a day, by text, until your test date. No app to install.