ownIT
Your student isn't bad at maths. They're missing something underneath the question.
Most tutors already know this. They just don't have a system that finds it fast.
Free, no sign-up. Paste a question or photograph one, and it works down to what is actually missing.
Why doing more past papers isn't working
Every A-level question sits on top of other things. Differentiation sits on indices. Indices sit on knowing what a power means. When a student gets a question wrong, the mark scheme says they got that topic wrong — so that's the topic they go back and revise.
But if the actual problem is two years further down, revising the topic changes nothing. They do it again, it still doesn't stick, and they conclude they're just bad at maths.
They're not. They can't see where it broke. Neither, from the outside, can you.
A real example
What examiners actually saw
June 2024, AQA Paper 1, question 10. Prove from first principles that the derivative of x³ is 3x². Here is what the examiners reported students doing:
Quite a few students expanded (x + h)³ as (x² + xh) + (xh)… because students did not collect like terms. AQA examiner report, 7357/1, June 2024
Every one of those students lost the marks on a differentiation question. But not one of them had a differentiation problem. They couldn't expand a bracket — something they were taught at fourteen.
This is what ownIT does with that. It asks a few short questions, going down one layer at a time:
A real run of the diagnostic, against the real skill map. The mistake it is answering with is the one AQA's examiners reported.
The second thing the examiners reported on that question was students using h = 0 instead of the limit. ownIT finds that one in two questions, because it's a different break in a different place.
What your student sees
How it works
-
Send the question you're stuck on
A photo of the page is best — typing maths out loses the powers and the fractions. Tell it what you tried, if you tried something.
-
Answer a few short questions
Usually about six. Multiple choice, and "I don't know" is always an option — it's more useful than a guess, and it isn't marked against you.
-
Get the first thing to fix — and what to go and do
Not a list of twelve weaknesses. One skill to start on, the reason it's that one and not something lower, and a line telling them what to practise — with an example and something to search. The first thing in the way, and it says so when there may be more behind it.
-
Climb back up to their own question
Four questions that start where the gap is and build, one step at a time, until the last one is the question they came in with — word for word, with nothing of the working given away. It's offered, not forced: they choose whether to take it.
-
Come back and run it again
After they've worked on it, the same question can be run from scratch. Nothing is remembered, so if the gap has closed the walk goes straight past it and finds whatever is behind it — which is the only way to know the work landed.
What you're left holding
A lesson plan, not a verdict
Knowing where the problem is only helps if you also know what it unblocks. So it doesn't stop at the gap: it gives you the order to work through, because each step is only reachable once the one before it holds.
Here is a second student on the same question — one further behind, whose trouble with brackets has taken algebraic fractions down with it:
That is the same chain the questions came down, read back the other way. Nothing about it is generated for the report; it is the route that was actually walked.
The obvious question
Couldn't you just ask a chatbot?
You can, and if you are good at prompting you will get something that looks a lot like this. Plenty of tutors already do. The difference is not the questions it asks — it is what happens to the answers.
When a model writes a diagnostic question, it also decides which option is correct. Nothing checks it. We measured that on this project: of the first 531 questions written, 55 had the wrong answer marked correct. Four in ten of the integration ones. Not one of them looked wrong.
That is the worst thing a diagnostic can get wrong, because it is invisible. A student answers correctly, is recorded as having a gap, and you spend a session teaching something they already knew. The report looks completely normal.
Every question here has been solved blind by a second model that was never shown the stored answer. Where the two disagreed, a third decided. 77 were retired.
The rest of the difference is the map. What differentiation rests on is written down, not improvised fresh each time, so the same student gets the same answer on Tuesday as on Friday — and when it gets something wrong, you say so once and it stays corrected.
Worth testing yourself: ask a chatbot for ten A-level integration questions with answers, and mark them.
It won't do their homework
It never gives the student the answer to the question they sent. That's deliberate, and it's the whole point — an answer gets them through tonight, and they're stuck again next week on the same thing wearing a different hat.
They can show the result to their teacher without it looking like they cheated, because they didn't.
It's open
Try it on a question now
I'm a maths and physics tutor. I built this because I keep finding the same thing in lessons — the question a student is stuck on is almost never where the problem is.
Bring a question one of your students is stuck on. A photo of the page works best. It takes about six questions and costs nothing, and there's nothing to sign up for.
If you're a tutor, add ?tutor=1 to the address
and you'll be asked at the end whether the diagnosis was right — and
what the real gap was, if it wasn't. That answer is the only thing that can
correct the map underneath this, and it takes one tap.
The part I actually need
Tell me where it's wrong
This is one tutor's theory about how maths failure works, with a tool attached. The theory is written down as a map of what rests on what — 296 links, and until last week not one of them had a reason recorded against it. I can't check that map on my own. A tutor saying "no, it was this instead" is the only evidence from outside my own reasoning.
So: if a diagnosis was wrong, if a topic you teach isn't covered, if a question wouldn't place, or if something just reads badly — that is the most useful thing you can send me. It goes straight to a page I read.