Digital SAT lessons and skill drills — 90 across 30 skills

The Digital SAT blueprint has 30 skills, and 90 micro-lessons across them — three per skill. The granularity is the point: "Boundaries" is a lesson, not a chapter of a punctuation unit you would have to sit through to reach it.

A lesson is a few cards carrying one idea, with a diagram where a diagram explains it better than a sentence would, and then recall questions. You do not advance by clicking next. You advance by answering correctly, which means a lesson you rushed does not register as a lesson you did.

That constraint is what makes proficiency mean something. Your number for each skill is computed from answers, and there is no path through the material that produces a high number without producing correct answers first.

A lesson teaches and then checks. A drill is only the check: question, answer, correction, next. No cards, no reading step, no unlock gate — fast enough that ten of them fit into the gap where you would otherwise open something else.

Questions due for review come first. After them, drills favour the skills where your recent results are moving away from your longer-term average — where practice is paying off, or where something is slipping — with a nudge towards skills you have barely tried, and each question says why it was picked: due for review, new for you, improving, or worth another look. A drill answer counts towards your skill estimate exactly as an in-lesson answer does, because it is the same question and the same evidence.

Every recall question you answer is scheduled for review with FSRS, an openly published spaced-repetition method. A wrong answer brings it back tomorrow; a right one pushes it further out — further still if you said you were certain, less far if you were not sure. The schedule aims for you to still remember nine in ten questions when they return.

Your daily mission says how many are due, and they are the first thing a drill serves, most overdue first. Everyone's reviews run on the method's published defaults until they have done a thousand reviews; after that, the timing can be fitted to your own history.

The skill numbers themselves are estimates from Bayesian knowledge tracing — a probability that you have learned the skill — and they are shown as ranges, because a percentage from six answers deserves visible uncertainty. Nothing is shown until five answers count.

Select a phrase in a lesson or a drill and you get one sentence explaining what it means. Not what the answer is — it will explain what a question is asking and refuse to solve it — and not a paragraph, because you were mid-sentence and want to get back to it.

It is deliberately fast rather than clever. The value of a gloss that arrives while you are still looking at the phrase is much higher than the value of a better gloss that arrives after you have moved on.

It is not available inside a timed exam, and that is structural rather than a setting: the exam runner does not import it. The honour code says that looking things up while you learn is good studying and is just not a ranked exam, and this is built on exactly that line.