Turn your lecture transcripts, slides, and teaching documents into board-style examination and practice questions, grounded only in what you upload, with a rationale for every option and everything ready to edit before it leaves your hands.
Add your teaching material on the left, set the blueprint, and the Augment Lab drafts questions here for you to refine, approve, and export.
Choose what to include, then pick a format.
Image items carry their image into Print and Word. Plain text notes the image by name, and GIFT embeds interpretation images while exporting hotspots as a placeholder to rebuild as a Moodle drag-and-drop marker.
A read of every question against NBME / NBOME item-writing guidelines. It flags cueing, weak distractors, unfocused stems, and reading-level problems, with a fix for each. It changes nothing on its own.
The Item Writer drafts examination and practice questions grounded only in the teaching material you provide. It is built for medical school faculty and writes to the style and standard of the NBME and NBOME item-writing guidelines.
Add any mix of PowerPoint slides, Word documents, PDFs, and pasted text. They combine into one knowledge base, and every question is written from that base. The model is instructed not to introduce facts your source does not support.
The Blueprint panel has a Learning objective field, above Focus areas. State one specific objective there and every question in that generation round is written to assess exactly it, taking priority over the more general focus areas. When you ask for several at once, the lab first maps out that many distinct angles on the objective, a different scenario, patient, and where the objective allows a different reasoning task, so the items assess the same thing without being near-duplicates. Leave it blank and the lab drafts its own objective for each question instead, which you can still edit afterward on the card.
The same spreading happens without an objective. With focus areas listed, the lab distributes items across them and adds distinct angles within an area when you ask for more items than areas. With neither field set, it maps distinct topics across the breadth of the source so a rich document does not return several variations of one concept.
The Your questions panel takes a file of questions you have already written, in Word, PowerPoint, PDF, or plain text, or pasted straight in. The lab reads it and transcribes each item word for word: your stem, your options, and your answer key if the file marks one, whether that is a bolded option, a line reading Answer, or a key listed at the end of the document. Nothing here joins the source material in panel 1, so your existing items are never used to ground other questions. Click any item in the list to see the full stem and options the lab read, so a bad transcription is caught before it goes anywhere.
From there the panel does two different jobs.
Guide the rewrite lets you say how you want your questions reshaped before anything is generated, so the variants come back close to what you want instead of needing heavy editing. Write one note that applies to the whole run, pick from the quick suggestions, or click any question in the list and add a note for that question alone. The lab follows your guidance ahead of the variant style, but never at the cost of the three pitfall rules below or the source material; if part of a note could not be followed, the card says which part and why. When you finish originals, guidance shapes only the rationales and teaching point, since your wording is never changed.
Items you brought in yourself carry a Faculty original label in the workspace so you can always tell your wording from the lab's, and the regenerate button asks before replacing one.
Add EKGs, imaging, histology, or clinical photos in the Image items panel. The lab reads each image and writes one question per image. Interpretation items show the image and ask a standard A to E question that requires reading it. Hotspot items ask the student to click a region: the lab proposes the correct area and you drag the box to fine-tune it, with a Test mode to check the target. Images travel with the question into the print and Word exports; in the Moodle GIFT file, interpretation images are embedded and hotspots export as a placeholder to rebuild as a drag-and-drop marker.
Three lapses come up again and again in drafted items, and each has a hard rule in every prompt, an automatic check after every draft, and a one-click repair on the card.
After each draft comes back, the lab runs a check before the item reaches your workspace. It counts the words in every option and looks for a key that stands out by length, scans the stem and lead-in for teaching language, and runs the cover-the-vignette test for real: it shows a second reader only the lead-in and the options, and if that reader can name the key with certainty, the item is a pseudocase. Anything that fails is corrected once automatically and marked Auto-corrected. Anything still left shows in a Pitfall check note on the card with Fix with AI, Recheck, and Keep as is. The note updates as you edit, and your own originals are checked and reported but never rewritten unless you ask.
A pseudocase, or pseudo-vignette, is a stem that wears the costume of a clinical vignette while the item underneath is recall: the patient, the age, the vitals, and the labs are on the page, but the case is decoration. The test is to cover the vignette and read only the lead-in and the options; if a prepared student can still answer, the vignette was never load-bearing. This matters because a pseudocase measures retrieval while appearing to measure clinical reasoning, which inflates apparent difficulty and misrepresents what your blueprint says you assessed.
The lab is instructed never to write one, at Foundational through NBME / NBOME alike, and to include only findings that constrain the answer set. Recall itself is not the problem; disguising recall as reasoning is. Where a point cannot support a load-bearing vignette, the lab is told to ask it plainly instead of adding a patient to the front. Check quality tests every item for this first, and the Adjust with AI box has a one-click repair for any that slip through.
Everything on a card is editable: the stem, each option, each rationale, the correct answer, the teaching point, and a learning objective. Two tools sit on each card. Adjust with AI (the wand) opens a box where you describe an overall change in plain language, for example that an image was misread, that the vignette should be shorter, or that the key should test mechanism; the lab rewrites the whole item, keeps what already works, and re-reads the attached image if there is one. Shuffle options reorders the choices so the correct answer moves through A to E rather than sitting in one place; new questions are already shuffled this way. You can also regenerate a single item from scratch, duplicate, reorder, or delete. When you are ready, export a print-ready exam and answer key, an editable Word file, a plain-text bank, or a Moodle GIFT file for your LMS.
Your questions and settings are kept in this browser so a refresh will not lose them. Uploaded PDFs are not stored between sessions; re-add a PDF if you want to generate more from it later.