AI that gives every student a method

The Smart DSE Generator turns a blank page into a full, exam-ready paper in a matter of seconds. Pick a subject, choose the paper you want to build — Reading, Writing, Listening & Integrated Skills, or Speaking — and the generator assembles a complete set of curriculum-aligned questions around it, from short comprehension checks to full extended-response tasks.

Before you commit, you get a live preview of the paper structure: total question count, estimated completion time, and how many sections it spans, so you always know what a student is about to sit down to. Every paper can be renamed, restructured, and fine-tuned — down to editing individual passages and swapping question formats — so it fits the exact skill you're targeting that week, not just a generic template.

Because every question type is built against real HKDSE exam formats and difficulty bands, students spend their practice time facing the same structures, phrasing, and pacing they'll meet on exam day — not an approximation of it.

Every paper a student completes feeds straight into their Performance Dashboard, turning raw scores into a clear picture of where they actually stand. A skill-grade radar chart breaks performance down by competency area — comprehension, grammar, structure, and more — so strengths and weak points are visible at a glance instead of buried in a single overall percentage.

Percentile comparisons show how a student is tracking against peers at the same level, while automated strength & gap analysis flags exactly which question types are costing the most marks. Instead of re-reading a whole paper to work out what went wrong, educators and students get a ready-made score summary by question type, so the next study session can go straight to the gap that matters most.

Open-ended "study more" advice rarely turns into an actual habit. The Study Planner fixes that by turning a goal — a subject, a target grade, an exam date — into an AI-generated, day-by-day schedule. Pick how many days you have and how many minutes a day you can realistically commit, and the roadmap builds itself around real life instead of an idealized version of it.

The plan isn't static once it's built. It prioritizes whichever topics a student is weakest in, balances the daily workload against upcoming exam deadlines so nothing is left to the last week, and adapts automatically as new performance data comes in from every paper and practice session — quietly re-weighting future sessions toward whatever still needs the most work.

Talk Anytime gives students a library of real conversation scenarios to rehearse before they matter — from everyday situations like ordering at a restaurant, to high-stakes ones like a job interview, a university admissions mock interview, or a full HKDSE Oral simulation. Each scenario runs as a real-time AI conversation with multiple characters playing distinct roles, so group discussions feel like an actual group, not a single back-and-forth with a chatbot.

Because there's no clock, no classroom, and no waiting for a partner to be free, students can practice as many times as it takes to feel ready — repeating the exact scenario that makes them nervous until it stops being nerve-wracking. For an exam format built entirely around live spoken performance, that kind of low-stakes, unlimited repetition is exactly what turns hesitation into speaking confidence.

Every question a student answers incorrectly — across every subject, every session — is automatically collected into their Individual Error Bank instead of quietly disappearing once the paper is marked. Mistakes are organized by topic rather than left as one long undifferentiated list, so patterns become obvious: three missed questions on quadratic factorisation read very differently from three unrelated one-off slips.

From that organized view, the system generates smart revision suggestions built specifically around each student's own mistakes, then delivers targeted reinforcement — fresh practice questions on exactly the topics that tripped them up — until the gap actually closes. It turns every wrong answer into the most efficient, personalized revision material a student has, rather than something they simply move past.

The AI Tutor is built to teach, not just answer. Instead of handing over a solution immediately, it opens with a warm-up question, asks the student to attempt the working themselves, and only offers a suggested answer as a safety net when they're genuinely stuck — Socratic, step-by-step guidance that develops real understanding instead of memorized steps that fall apart on a slightly different question.

Every explanation stays curriculum-aligned to what the student is actually being examined on, and the support adapts to how that individual student is progressing — moving faster through concepts they've clearly grasped, slowing down and re-explaining wherever they're still shaky. Because it never sleeps, that same patient, personalized tutoring is available 24/7, whether a question comes up during homework at 4pm or last-minute revision at midnight.

Smart Grader marks a single question straight from a photo in about a minute — no quiz or assignment needed. Choose Equation mode for maths or Writing mode for essay-style responses, then frame the question in the viewfinder and capture or upload it. The AI reads the question back so you can check and correct anything it misread before moving on.

Next, capture or upload a photo of the student's working — handwritten class work, a worksheet answer, or something on a whiteboard — and Smart Grader marks it against the question in roughly a minute. The result shows a score out of the question's total marks, plus a step-by-step marking scheme that breaks down exactly where marks were earned or lost.

Every result also flags specific weak spots, keeps both the question and the student's response editable in case a photo was misread, and saves each grading under Recent so nothing gets lost between one-off checks. It's built for the moment marking needs to happen right there in the classroom, not after work has been typed up or formally submitted — a different job from in-quiz auto-grading, which marks assignments students submit through the platform itself.