Read a class set in five minutes with one prompt

Read a class set in five minutes with one prompt

You've got the class set on the kitchen bench and it's Friday. You're not going to mark every one tonight and you shouldn't. What you actually need before Monday is a lens, not a grade: which errors are common, which are one-off, and where to spend the first fifteen minutes.

The prompt

You are a strict, experienced [SUBJECT] teacher analysing a class set. I am going
to paste in short excerpts from [NUMBER, usually 6 to 10] students' work on the
same task. Do not mark them. Do not give me a grade per student. I want a
pattern-level read of the class.

Task context:
- Year level: [YEAR LEVEL]
- Subject: [SUBJECT, and if VCE/VCE-VM/ACARA v9.0 outcome, name it]
- Task: [ONE SENTENCE describing what students were asked to do]
- Success criteria I care about most: [2 TO 4 CRITERIA, plain language]

Student excerpts (each labelled with a student pseudonym so I can find them
later):
[PASTE EXCERPTS HERE, one per student, 1 to 5 sentences each]

Return exactly this, and nothing else:

1. SHARED ERRORS. List the 2 to 4 mistakes or misconceptions that show up in
   more than one student's work. For each, name (a) the misconception in one
   sentence, (b) which students show it and how you can tell, (c) whether it
   looks like a knowledge gap, a skill gap or a task-comprehension issue.

2. ONE-OFFS. List up to 3 errors that only appear in one student. Name the
   student and what to do about it (a one-line intervention).

3. WHAT'S GOING WELL. One or two things the class is actually doing well.
   Be specific and evidenced. Do not pad.

4. MONDAY'S MOVE. Give me 2 or 3 whole-class teaching moves for Monday's
   lesson that would address the shared errors, in priority order. Each move
   is one sentence and describes what I actually do, not a principle.

5. DO NOT DO. One thing I should NOT do based on this data (e.g. don't
   re-teach a topic they already have, don't set the extension task yet).

Australian English. No em-dashes. Be blunt where the evidence is blunt.

What it does well

It refuses to grade individual students, which is the trap most AI marking prompts fall into. You get a pattern-level read that tells you where to spend Monday's first fifteen minutes, not a false-precision score you'd have to double-check anyway.

Where it falls over

It's only as good as the excerpts you paste. If you feed it the strongest and weakest and nothing in the middle, the "shared errors" section will be misleading. Grab responses from across the range, and if your task was a long piece, paste only the sentences most relevant to the criteria (the model can't read minds about what "good" looks like).

Try it like this

• For Year 8 science: paste in the "explain why" sentences from a phase-change worksheet, ask for shared misconceptions about energy versus temperature. Then use "Monday's move" as your do-now.
• For VCE English: paste the topic sentences from ten essays, ask for shared errors in framing the contention. You'll see the whole-class thesis problem in about a minute.
• For primary maths: paste the working (not the answer) from a set of word problems, ask which step students are getting stuck at. Nine times out of ten it's the translation step, not the arithmetic.
• If you only have five minutes: just paste six responses and skip the one-offs section. The shared errors alone are worth the coffee.

Pair this with Wednesday's report-comment prompt at the end of term and you've reclaimed most of the weekend.