Loading...
Co-Teacher is the AI assistant for teachers. Photo-grade handwritten work from a phone shot. Author Blueprint exams by Bloom’s percentage. Friday digest with action cards. Every output reviewable. Teachers ship; the AI assists.
Most “AI for education” pitches replace the teacher. We don’t. The teacher is the relationship — with the student, with the parents, with the room. AI can do the drafting; it can rank what to attend to first; it can compress an hour of grading into fifteen minutes. But the judgement that turns those drafts into decisions is the teacher’s.
Every surface in CurioPilot is built around that asymmetry. The Friday digest proposes six action cards; the teacher chooses which to pursue. The grading queue drafts a rubric breakdown; the teacher confirms or overrides. The parent message is drafted; the teacher edits and sends. The override path is one click — never buried, never nudging the teacher toward “accept default”.
This isn’t a feature. It’s the shape of the product. We log who drafted, who edited, who shipped — and that audit trail is something teachers can hand to a parent, to a department head, to a regulator. Visibility belongs to the teacher.
Every AI output is a draft. The teacher is the publisher. That principle is wired into the product, not the marketing.
Twelve minutes between opening the laptop and having Monday planned, two students intervened with, and a parent conference booked. Here’s how it actually goes.
“Friday afternoon used to be 90 minutes of grading. It’s the digest now — read it, edit two cards, ship them.”
Time-saved figures are honest beta-data numbers from our pilot cohort. We’ll update them as the data updates.
Six Action Cards per week, ranked by what'll move the needle most. Each card cites the data: how many students at risk, on which topic, with a recommended next step. Read it Friday afternoon, plan Monday in fifteen minutes.
When a student spikes in risk mid-day — failed three quizzes in a row, dropped two Bloom's levels, abandoned an essay halfway — the teacher gets a toast notification. Dismissible, escalatable, or routed to an intervention. Nothing is broadcast to the whole class.
Report-card comments, intervention notes, parent-message drafts. AI generates the first pass grounded in the student's actual data; the teacher edits and approves.
Sarah has shown strong growth in fractions, particularly in simplifying to lowest terms. Next term, we’ll focus on…
School admins see the same Action Card model triaged across the whole school. Department heads can spot a misconception running through three sections of Year 5 Math; a principal can see which classes are humming and which need backup.
Every AI-generated grade or rubric is editable. The override path is one click — not a buried flow that nudges teachers toward accepting the default.
Snap a photo of a student’s notebook. The platform OCRs the handwriting, runs the output through the same rubric the AI generator used to author the question, surfaces partial credit on multi-step problems, and lets you confirm or override before the grade lands in the gradebook.
Author a unit exam the way your curriculum coordinator already writes it: “30% Apply, 40% Analyze, 30% Evaluate across these five topics.” A Hamilton-rounding allocator fans the right counts into the right buckets; the question generator fills each bucket — dedup-aware, mastery-aware, reviewable. Auto-publish is OFF — half-built blueprints never accidentally become live activities.
Exam Prep (mastery-gated pacing). For pre-board prep across a whole grade, Exam Prep paces practice to the exam date. Audiences resolve server-side from divisionIds — never trusted from the client. See the Exam Prep brief.
Mind Mirror’s teacher surface. Twelve traits — Phoenix Spark, Wayfinder, Mosaic Mind, and nine more — surfaced as four positive bands. No leaderboard. No raw scores. No “weak” or “struggling” copy anywhere, even on your view.
Aggregate participation from Boss Arena and Party Hub flows here without surfacing per-student numbers. You see the shape of the room; you don’t see a ranking.
Activities embed inside your existing LMS via LTI 1.3. Grades pass back to the LMS gradebook through a durable queue and a nightly worker, so retries are crash-safe and your students never see a missing grade due to a transient network blip.
Field-level encryption on client secrets and SP signing material. Every passback is audit-logged in TraceLayer.
The OCR pass returns a confidence score per recognised token. Where confidence drops below threshold, the teacher sees the raw photo alongside the AI's interpretation, with the uncertain tokens highlighted. Override is one click; nothing lands in the gradebook without teacher confirmation. Image moderation is fail-open on student-upload paths so a child's submission isn't blocked by a moderation timeout.
Blueprint exams let a curriculum coordinator author by Bloom's percentage — "30% Apply, 40% Analyze, 30% Evaluate across these five topics." A Hamilton-rounding allocator fans the right counts into the right buckets, then the question generator fills each bucket — dedup-aware, mastery-aware. Auto-publish is OFF by design so half-built blueprints never accidentally become live activities.
Yes. Grade passback runs on a durable queue with a nightly worker, so retries are crash-safe and students never see a missing grade due to a transient network blip. License-gated under lti-integration (Standard + Campus).
Yes. The digest is a Friday email plus an in-app card; both can be dismissed, snoozed, or unsubscribed per teacher. Action Cards never auto-execute. Dismissing the digest dismisses the cards — it doesn't grade anything, message anyone, or assign anything on its own.
The teacher. Always. AI drafts a rubric breakdown and a suggested grade; the teacher reviews, edits, or overrides — one click. The audit log records who drafted, who edited, and the final grade is attributed to the teacher in the gradebook. Schools can also turn off AI grade-suggestions entirely at the class or campus level.
Alerts fire on aggregate signal, not individual events. A single failed quiz is noise; three failed quizzes in 40 minutes after two Bloom's-level drops is signal. Each teacher can tune thresholds, mute alerts during teaching periods, or route them to a co-teacher or counsellor instead.
No. Co-Teacher drafts and ranks; the teacher decides. Every output is reviewable, editable, and overridable. We don't claim the AI knows the student better than you do — it knows the data better than you have time to read. The judgement is yours.
Co-Teacher is opt-in per teacher. Schools can deploy CurioPilot's gradebook, attendance, and curriculum tooling without enabling Co-Teacher at all. If a teacher disables it later, their account stops generating drafts immediately and the existing audit logs stay intact.
30-minute walkthrough. We’ll run a sample Friday digest against an anonymised class and show the override + audit flow end-to-end.