For teachers · in development
Marking 120 sets of working, by hand, every week.
Answer keys check the box at the bottom. They can't tell you that eleven students in your third period all dropped the negative on the same distribution step — which is the thing you'd actually change your lesson for.
No account, no card. One email when there’s something to show you.
What already works
The engine behind this reads handwritten math from a photo and checks it line by line — not by comparing a final answer, but by asking whether each step follows from the one above it. When it doesn't, it names the line and the rule that was skipped. Students use it today to check their own work before they submit.
Everything below is the same capability pointed at a stack of papers instead of one page.
What we think the class-set version does
- A stack, not a page. Photograph or scan a set, and it splits it into individual submissions rather than making you crop thirty times.
- Method, not just answers. Marks the first invalid line in each student's working — so partial credit has something behind it, and "right answer, broken method" stops slipping through.
- Error patterns across the class. The part we think actually matters. Not thirty scores, but "nineteen of thirty lost it at the same place" — which is a lesson plan, not a gradebook entry.
- You keep the pen. It proposes; you decide. Nothing is submitted to a gradebook on your behalf, and you see the transcription before you see a verdict.
Things we'd want to get right
We'd rather raise these than have you discover them:
- Student data. A photo of a worksheet carries a name in a child's handwriting. Anything we build for classroom use needs the data agreement and retention terms your district will ask for, before it ships — not after.
- Being wrong. A grading tool that confidently mis-marks is worse than no tool. Which is why it shows what it read, flags what it's unsure of, and never files a mark without you.
- Handwriting that's genuinely hard. Some of it is. We'd rather say "can't read this one, look yourself" than guess.
Why we're asking instead of building
We've built for students for a year. Students use a tool the night an assignment is due and then disappear until the next one. Teachers grade every week, all year — which is a completely different shape of product, and not one we should assume we understand.
So: if you'd use this, say so. If you'd use it only if it did some specific thing, say that instead — it's more useful than the click.
Nothing to buy here, and no product yet. We already do step-level checking of handwritten work for students, one problem at a time — the question is whether the class-set version is worth building. We're counting teachers before we commit, and if we decide against it we'll tell you rather than leave you on a list.