Uses real tools
Looks things up, reads a document or handout section by section, checks notes — grounded in an actual source, not a plausible-sounding guess.
Not a chatbot that guesses from memory. A teaching assistant that reads a real document, checks a real source, and hands off to a stronger model the moment a question gets hard — verified by a second, independent AI before an answer ever reaches a student.
An ordinary chatbot answers from what it remembers. This assistant is built to actually look things up — and to catch itself when something goes wrong along the way.
Looks things up, reads a document or handout section by section, checks notes — grounded in an actual source, not a plausible-sounding guess.
A watchdog tracks real progress through a multi-step task, not just a flat timer — a quietly stuck or failed attempt is caught, never mistaken for a finished answer.
Before an unfinished plan or a long, sourceless answer is treated as final, a genuinely different AI model reviews it — verified, not just generated.
Starts with a fast, economical model; if that one struggles, the same question is retried on a stronger model within the same request — the student just gets a good answer.
Can: look things up, read a document or handout front to back, check notes on request — real, bounded tools, not an open-ended ability to act on anything it imagines.
Can't, by design: run arbitrary system commands, or reach into a student's messages or personal accounts — left out on purpose, not yet gotten around to. What it's allowed to touch is a short, explicit list, not "everything the model can imagine."
Every answer's real cost is tracked and billed on actual usage, not a flat guess — and the whole thing has been tested end-to-end on a production-matching server before ever reaching a real classroom.
Let's talk about where a verified, tool-using assistant fits into your classroom.