300 questions. Two separately timed papers. No negative marking. Half of them right and you are through — and nobody you are sitting with can take that from you. Everything here is pointed at that line and nothing else.
Not a screenshot and not a mock-up — a real question from the bank, playing the loop you will actually perform: eliminate, commit, flag, then take the answer apart.
Which tuning fork frequency is the standard choice for performing clinical Rinne and Weber hearing tests at the bedside?
A faithful NBE sitting: Part A and Part B on separate 150-minute clocks, a question matrix, flagging, option strike-through, and an image inspector with zoom, pan, contrast and a measurement reticle. The clock lives on the server — reloading the tab cannot buy you a second.
A projected score weighted by blueprint share, with each subject pulled toward the guess level in proportion to how little you have been tested on it. It climbs on evidence, not on one lucky mock.
Every question you get wrong is queued automatically and returns when you are about to forget it. FSRS, not a fixed ease multiplier.
Every wrong option explained, and what condition it WOULD have been right for. That is where the marks are.
Fast-and-wrong is a different problem from slow-and-wrong. We separate them, because the fix is different.
Short, sourced, cross-linked to the questions they answer. Printable, searchable.
Each shows the formula it used and the paper it came from. Learn the derivation, not the output.
Eligibility, application, exam day, registration after passing — each naming the authority that decides it.
A past paper can tell us which concept was tested. It cannot give us the question — so we do not take it. Every item is written from primary sources around a fresh patient, and the pipeline refuses to publish anything that lacks a source, an in-house reviewer, or an explanation for every distractor.
Written from primary sources around a novel patient. Where a past paper suggested the concept, only the year is recorded.
Schema-enforced: every distractor explained, every published item sourced and reviewed in-house (independent faculty review is being rolled out).
Answer positions are re-lettered deterministically, so nobody can pass by learning that the answer is usually A.
Only reviewed items reach candidate-facing tests. Coverage against the blueprint is tracked in the open.
Including the ones with answers that do not flatter us.
No. Every question should be answered. A blank is a guaranteed zero; a guess on a four-option question is worth 0.25 marks on average, so thirty blanks costs you roughly seven or eight marks — often the width of a failure. The submit dialog tells you exactly what your blanks are costing before you confirm.
Start free and see where you stand. When you commit, one payment covers everything until the exam you are sitting — and if you do the work and just miss, the next cycle is on us.
Find out where you actually are.
This is live now — and during the beta the image inspector and weakness heatmap are switched on too.
Everything, until the exam you are sitting for.
Coming soon. One payment when Pro launches, not a subscription — join the founding waitlist below for launch pricing.
Do the work and miss 150 by 15 marks or fewer (a verified 135–149) and your next full cycle is on us.