The best-evidenced learning techniques — recall, spaced repetition, explaining your reasoning, connecting your learning — take effort. There are no end of question banks, but Mosaic offers much more: a suite of tools for GP registrars built around the evidence for what works in learning, mapped to the RCGP curriculum, and free to use.
Get startedMosaic is in pilot. You'll need an invite code from your training programme to create an account.
What's inside Mosaic?
Every tool in Mosaic exists because a specific learning technique is backed by evidence. Use them together, or pick whichever suits you today.
Spaced retrieval. Testing your knowledge beats re-reading, and spacing out your reviews beats cramming. Mosaic's flashcard app combines the two best-evidenced techniques. Your confidence in the knowledge — and the FSRS algorithm — decide when each card comes back for testing, and you can create banks of cards and share them with the community.
Generation and self-explanation. Writing an MCQ, including explanations of why each option is plausible yet wrong, helps embed and refine your knowledge. AI feedback checks the quality of your question and offers feedback.
Connection and prediction. New material sticks when you make predictions, attach new knowledge to things you already know, or relate it to your clinical practice. Mosaic's connection notebook offers a simple way to record and demonstrate these connections across the RCGP curriculum.
Learn from your peers. Browse flashcards and review AKT questions other registrars have published, upvote them, see how they defended each option, and have a discussion, all in the app.
Learner-produced visualisations. Drawing concept maps, which summarise everything you know about a topic, is a powerful way to learn: it forces you to select what matters and organise how diagnoses, mechanisms, drugs and decisions relate.
The same write-then-review loop from Mosaic's AKT writing app, applied to consulting. Create scenarios for SCA prep with AI feedback.
The approach
The most popular study techniques — rereading, highlighting, and watching — feel productive but rank at the bottom of the evidence table when we look at what actually works in learning. The techniques that work are effortful and generative: you make something, test yourself against it, and find out what you actually know. That principle runs through every Mosaic tool.
Struggling to recall the back of a card, writing a plausible distractor or predicting an answer before you know it, is not a sign that the technique is failing. The effort of generating from memory is what strengthens it.
Predicting before you read, defending an answer that turns out to be wrong, getting a card wrong twice in a week: each one surfaces a learning challenge while there's time to fix it.
New knowledge holds best when it attaches to what you already know. Mosaic keeps pointing back to your own patients, your education sessions and the RCGP curriculum, so what you learn is organised the way you'll need to use it. Your personal dashboard shows how all your learning activity links together.
How it works
Mosaic isn't content to get through. It's a cycle you run, in whatever order suits the week you're having.
Write an AKT question on a topic you're working on, make flashcards from it, or log a connection from something you've learned this week. Everything maps to the RCGP curriculum and topic guides.
AI review checks your question's clinical accuracy, construction and distractor quality — option by option, against your own stated reasoning. If your intended answer is wrong, it says so. Retrieval works better with a corrective loop than without one. Throughout, we remind you that AI can hallucinate, make errors and be out of date. Checking the accuracy of the AI feedback can be part of your learning.
Flashcards come back just before you'd forget them, spaced further apart each time you get them right. Spacing is decided by the scheduler, guided by how confident you are on each question.
The curriculum dashboard shows which topics your questions, cards, and connections actually cover. The gaps tell you what to create next — and the loop starts again.
The evidence behind the approach
In the most cited review of study techniques (Dunlosky and colleagues, 2013), only two techniques earned the top utility rating: practice testing and distributed practice. Mosaic is built around both, plus a small set of moderately evidenced techniques that suit clinical learning particularly well.
Bringing information to mind from memory strengthens it more than simply re-studying it does. The effect is large and has been replicated across decades, content types and settings, including medical education (Roediger & Karpicke, 2006; Larsen et al., 2009).
The same study time spread over weeks beats the same time massed into one sitting (Cepeda et al., 2006). Spacing of 10–20% of the time before you will be tested on the information seems most powerful. Mosaic's flashcard scheduling applies this automatically, with intervals matched to how long you need to retain the material.
Explaining why an answer is right — and why each alternative is wrong — builds the kind of understanding that transfers to new cases (Chi et al., 1989). This is why Mosaic approaches question writing this way.
Material linked to what you already know is retained better (Ausubel, 1968), and predicting before you learn improves recall of the right answer even when the prediction was wrong (Kornell et al., 2009). The connection book is built on both.
But… much of this evidence comes from laboratory and classroom studies. Replications in postgraduate medical education are encouraging but are fewer. We think building on the best available evidence beats building on habit — but no app makes learning effortless, and we won't claim this one does.
Built with care
Mosaic uses AI to support learning, not to replace clinical judgement. We're upfront about what it can and can't do.
Every review comes with a reminder that AI feedback should be verified against current guidelines. This is a learning tool, not a clinical reference. Checking the AI's accuracy is a powerful way to learn.
Feedback cites NICE, the BNF, RCGP curriculum guides and other UK sources by name and number rather than generating links. This reduces, but does not remove, the risk of an inaccurate reference — check anything you plan to rely on.
The AI is prompted to flag uncertainty rather than guess, and some questions have genuinely debatable answers. It will still sometimes be wrong, which is why every review carries a reminder to check current guidance.
Database hosted in London (AWS eu-west-2). Personal data is not shared with third parties. Questions you write are private unless you choose to publish them to the community.
Create your account, then start wherever suits you — a flashcard deck, your first question, or a connection from this week's clinic.
Go to Mosaic