How does Arf teach?
Askarf is not an "answer machine." It blends the 2,400-year-old Socratic method with modern AI: with the right questions, the student finds the answer themselves.
Your student taps their own profile and speaks into the microphone. No typing, no getting lost in menus. If the question is a photo, they show it to the camera Arf reads it.
As you talk, the steps come to life on the board: equations, shapes, arrows. At each step Arf asks a single guiding question it never turns into a lecture video that explains the solution.
x = ?
Four rungs: nudge → concept → similar example → single step. Each rung helps a little more, but the final answer always stays with the student. Asking for a hint doesn't lower the score.
When the session ends, a clean summary appears in the parent dashboard: what was solved, where they got stuck, what's improving. Arf remembers what it learned too the next lesson starts right where it left off.
3 questions solved · 1 hint · negative-sign error improving
Behind the scenes: three-layer assurance
Every Arf sentence passes through three checks before it reaches the student.
A second model independently checks every step the student takes and hands Arf a note on "where and how it went wrong."
Mistakes are mapped to known misconceptions, so the hint starts at exactly the right step.
Every sentence is scanned separately so the answer never slips out of Arf's mouth; if it does, the sentence is regenerated.
I ask the right question you find it."