Guide, then reinforce.
Most “safe” AI tools either answer everything or refuse everything. Tutor mode does neither.
Refusing frustrates students. Just giving answers teaches nothing.
When every question gets a question back, students give up and go to tools with no guardrails. When every question gets an answer, they learn nothing. Tutor mode finds the middle: it gives sourced help when that helps, then makes sure the student does something with it.
Help that fits the moment.
Ask
What do you know? What have you tried?
Hint
Points to the right idea or source.
Example
Shows a similar problem worked through, not your own.
Explain
Gives the answer or explanation, with a source.
Check
A short task that uses what you just learned.
It starts at the lowest step that fits. A quick fact starts near “Explain.” A maths problem starts at “Ask.” If a student is stuck, it moves up one step at a time. It never skips to the finished assignment.
Four everyday moments.
The assignment stays the student’s.
- Never writes the essay or the answer to an assigned problem
- Never fills in a quiz the student is taking
- Never summarizes the student’s own assignment text for them
It will happily summarize approved library sources, with citations, and then ask what the student noticed.
Designed around how learning works.
Tutor mode uses well-established learning techniques: recalling information, spacing out review, studying worked examples, and getting feedback. We designed it around them. Whether it works in your school is something we’ll measure together.
Tutor mode is in design and will be tested in the pilot.
Help shape it.
We’re working with a small group of schools to build and test Trailpost. Pilot schools help decide what success looks like.
Join the pilot