Plan a lesson in one sentence. Know who's stuck, and where.
Copilot builds the lesson from your own question bank. Students practise with a buddy that never gives answers — then every gap comes back to you.
Most common slip
Sign flips when isolating x (38% of class)
The problem
You only find out after you've marked.
Planning eats your evenings.
Every lesson plan means hours picking questions, grading difficulty, and weighing up where each group is. Same again next week.
A score never tells you why.
A 5 out of 10 says the student is struggling. Which concept, what kind of mistake, how many times it repeated — the score stays silent.
General-purpose tools just hand over the answer.
The student asks a chatbot, copies the solution, submits. Homework: done. Learning: not.
How it works
You assign. They practise. It reports back.
- Step 1
Plan a lesson in one sentence
Type one line: “Quadratic equations, class 9A, upper-middle group.” Copilot pulls from your own standardised question bank, orders the problems along the P1→P4 arc, and drafts the groundwork session. You review, edit if needed, and only then publish.
- Step 2
Students practise with D-Friend
Not a tutor re-explaining the lesson. A study buddy sitting beside them, following their approach, raising doubts instead of pronouncing errors — and never giving the answer away.
- Step 3
A report that goes down to the mistake
After the session you get a report: where the class stands, who needs help, and why — at the level of concepts, not scores.
After every session
Not just a score. A diagnosis.
- Class average
- 6.8
- Attended
- 32/35
- Problems attempted
- 124
How the class split
Top 3 sticking points
- 1Sign flips when isolating x38%
- 2Forgetting the Δ > 0 condition25%
- 3Misremembering the root formula19%
Worth a check-in
Gia Bảo
Flips the sign when isolating x, in both problem 2 and problem 3.
Thu Hà
Right answer on problem 3, but got there by testing numbers.
Minh Khôi
Stopped at problem 2, hasn't come back after three attempts.
Suggested for next session
Revisit: moving terms across, and the condition for two distinct roots
Draft the next lessonThe whole class at a glance
Class average, who turned up, how many problems were attempted. One look, ten seconds.
Three groups
Solid · struggling · didn't finish. You know the shape of the room before the next period starts.
Top 3 sticking points
The problem types and concepts the class tripped on most, ranked by how often they came up.
Who needs you, and why
Not “student A: 4.5” but “student A: flips the sign when isolating x, in both problem 2 and problem 3.”
A suggested next step
Move on, revisit, or push further — and one click to draft that lesson.
Every report ends with a button. One click and the next lesson is drafted, tuned to exactly what the class just tripped on.
You're in control
The system drafts. You decide.
Nothing reaches students unreviewed.
Every lesson plan and every extra exercise stays a draft until you have read it, edited it, and pressed publish.
Questions come from your bank, not invented.
You upload your papers in bulk and the system breaks them into structured exercises. Copilot only selects and sequences from that library — it does not write practice content of its own.
Built for the Vietnamese curriculum.
Vietnamese exam structures, Vietnamese conventions for writing up a solution, Vietnamese as the first language — not a translation layer.
Why it's different
Right answer, wrong method — we catch that.
Most tools mark in one dimension: correct answer, well done. But a student who reaches the right answer by the wrong route carries that route into the exam — and it fails them on the harder question.
D-Friend reads the method and the result as two separate signals. Right result, wrong path? The system records it, you see it in the report, and the study buddy nudges the student into looking at their own working again.
A typical tutor bot
“You forgot to check the case x = 0.”
D-Friend
“Huh, I did it that way too and got stuck right about here… what did you do at this step?”
School leaders see the whole school.
Learning data rolls up by class and by year group — which concepts a whole cohort is weak on, which teachers could use more support. Register for a school-level pilot and talk to us directly.