Human in the loop

Review queue

Evidence first, the AI's idea second, your decision always. A short list with an end. Not an endless feed of alerts.

Scans the system isn't sure about

The system won't guess on a messy scan. A misread mark and a real mix-up are very different, so it asks you.

Low-confidence scan — step 2 unreadable

Ishaan Banerjee Human review required

Handwriting parse confidence 0.41 on Ishaan's worksheet, item 3. The machine can't tell a careless slip from a real misconception here, so it won't guess — your read decides.

MATH.FRAC.EQUIV.01 · Recognise and generate equivalent fractions

Re-scan suggested

Diya Patel Human review required

Two answer regions were left blank or scanned at an angle. Re-scan when convenient — no rush.

MATH.FRAC.MEANING.02 · Represent fractions on a number line and as part of a collection

Your call to make

The tutor only checks the mechanics. It never grades thinking. Meaning, argument, and voice are yours to judge.

Judgment node — does the idea land?

Aarav Sharma Human review required

Tense flips were auto-flagged (mechanical). Whether Aarav understood and carried the story is yours to decide — the tutor flags, it never scores thinking.

ENG.WRITE.NARR.05 · Develop a narrative with a clear, well-carried idea

Content to approve

Put together by trusted tools. A quick look before it reaches a child.

Support worksheet awaiting approval

Human review required

AI-differentiated support sheet for today's equivalence group. Nothing reaches a child unreviewed — approve, edit, or reject.

MATH.FRAC.EQUIV.01 · Recognise and generate equivalent fractions