Coid

Coid's whole world is the material uploaded for the module you are in: your books and your notes. It is not searching the web, it cannot read another module's material, and it is not answering out of the general knowledge of the model underneath it.

That is a deliberate limit, and it is what makes the thing useful. A general chatbot will answer any question about medicine; it will not answer the question your department is going to ask, in the words your department uses, from the book your department set.

One — Type it, or say it

The composer takes a typed question or a dictated one — hold the microphone and talk, and your words arrive as text you can edit before you send them. Ask in whatever language you think in.

Two — It finds the pages first

Before it writes a word, Coid looks through your module's material for the passages that bear on what you asked. Those passages, and not its memory, are what it then answers from.

Three — The answer arrives as it is written

It streams in rather than appearing all at once, so you are reading while it is still going. Underneath it, the pages it used.

You can also start from a question you got wrong. Asking Coid about it from inside the exercise carries the question across with you, so you do not have to describe what you were looking at.

Sources sit under the answer as a row of pages you can open — never as a small number buried in the text. Tap one and the book opens at that page, in the app's own reader, with everything around it intact: the paragraph before, the figure beside it, the rest of the chapter if you want it.

Each distinct page appears once. An answer built from three paragraphs on the same page gives you one page to open, not three identical buttons — the list is somewhere to go, not a count of how hard it worked.

How much you want, and whose words you want it in, are two separate choices — so you can have the short version in your book's own terms, or the long version explained plainly, instead of picking one bundle and living with it.

An answer is not always a paragraph, either. Ask for a comparison and you get a table; ask how something branches and you can get a map you can follow; a long answer arrives with headings you can scan rather than as one block to wade through.

Four limits, each with the reason. A tool that quietly does less than it claims is worse than one that tells you where it stops.

• Anything your module has no page for

  It answers from your books. Ask about something they do not cover and it says so rather than reaching for what the underlying model happens to know — which is exactly where a confident, wrong, unsourced answer would come from.

• Another module's material

  Coid sees the module you are in and nothing else. Switch modules and it switches with you. This is the same rule the rest of the app follows, and it is why the answer you get is the answer your year is examined on.

• Medical advice about a real person

  Coid is a study tool built on teaching material. It is not a clinical decision aid and it has no business anywhere near a patient — yours, your family's, or your own. For that you need a doctor, not an app.

• A source it cannot show you

  If it cannot point at the page, it does not dress the answer up as though it could. You will get the answer without the sources rather than a citation invented to make it look grounded.

There are two models behind Coid. Coid Lite is where every conversation starts — quick, and enough for most of what you will ask. Coid Pro comes with Clinicoid Pro; you reach for it from the switch at the top of the chat when a question is worth it.

Coid Pro is a per-conversation choice, not a setting you leave on. Open the app tomorrow and you are back on Coid Lite — so nobody spends a week's allowance on questions that did not need it without meaning to.

Free and Clinicoid Pro get the same Coid Lite, with the same features. A subscription adds Coid Pro and more room — a bigger allowance and no daily ceiling. Both models draw on it, Coid Pro faster, and Pricing sets out what the free plan includes, and what Clinicoid Pro raises it to.

Conversations are filed by module, so the chats from the module you are studying are the ones you see. They are kept on your account so you can come back to them, and you can delete any of them yourself, at any time, without asking us.

Answering a question means sending your question, the passages it matched from your books, and the recent turns of that conversation to the model that writes the answer. Your name, your email address and your account are not part of that. What is stored, for how long, and how to remove it is set out in the Privacy Policy.