Project Record
Problems.cc
Adaptive AI training platform that models a student's knowledge over time, selects the next useful problem, and evaluates written reasoning—not just final answers.
Executive Summary
Problems.cc turns mathematics and computer science preparation into an adaptive process. Instead of presenting a flat problem bank, the platform analyses a student’s level, solution history, typical mistakes, and progress dynamics to select the next most useful problem, set, or past paper.
The system does not only check answers. It builds a long-term model of knowledge across topics, surfaces gaps, evaluates written solutions step by step, and personalises training at the individual level. That makes it suitable for academically strong students, olympiad preparation, and complex educational routes where generic EdTech products are too shallow.
The platform combines daily practice infrastructure (dashboard, quests, streaks), a hierarchical knowledge map, interactive library lessons, competition past papers, gamified daily puzzles, and written-reasoning feedback. Public entry points allow discovery without registration; signed-in users get persistent progress, diagnostics-linked recommendations, and a full training history tutors can review.
Problem
Olympiad-style preparation fails when practice is unstructured. Students bounce between random worksheets, past papers, and apps that only mark final answers. Tutors lose visibility into where reasoning breaks down. Parents cannot tell whether progress is real or just repeated easy tasks.
The core challenge is not content volume—it is sequencing. A useful training system must know what a student has attempted, which topics are weak, which mistake patterns recur, and what difficulty comes next. Written solutions matter as much as multiple-choice scores, especially for competitions that reward clear argument.
Solution Path
The platform is organised as a set of connected modules around a single learner model:
- Onboarding and diagnostics establish an initial skill profile and route new users into appropriate practice.
- Knowledge map structures the curriculum as a hierarchy of topics, each linked to problems and progress state.
- Adaptive practice selects the next problem, worksheet, or past-paper section from history and topic performance.
- Written-solution feedback analyses reasoning chains and flags logical gaps—not just answer correctness.
- Engagement layer (daily games, quests, streaks, leaderboard) keeps regular practice sustainable between lessons.
Public surfaces (games, problem of the day, catalog pages) lower the barrier to entry; the signed-in experience holds the full training record.
Implementation
Dashboard and quests
The personal area sets daily goals—games, practice sessions, library lessons—and tracks quest completion. Streaks and progress indicators give students a clear next action rather than an empty problem list.
Knowledge map
Topics are arranged hierarchically so navigation follows knowledge areas, not arbitrary problem IDs. Each branch links to specific problems and records what the student has solved, attempted, or skipped.
Library and interactives
Twenty-nine ready lessons span Primary through Senior stages. Each lesson page combines short theory, interactive sandboxes, and immediate practice problems—usable by teachers in class or by students preparing independently.
Daily games
Cryptarithm, Magic Square, and Summit run in the browser with UK-specific date variants. Guest mode supports casual play; registered users contribute to completion analytics and streak tracking.
Practice and past papers
Practice mode supports quick entry by problem code or curated selections, with worksheet generation for classroom use. The past-paper archive covers major competitions with admin tooling for inventory (JMO, Cayley, UKMT, AMC) and award boundaries by year.
Profiles and diagnostics
Public profiles expose solve history and rankings. The diagnostic flow links test results to personalised dashboard recommendations so new users do not start from a generic default.
Trade-offs
Written-solution feedback is more expensive to run than auto-grading but essential for olympiad training where process matters. The knowledge map requires ongoing curation as topics and problem links evolve. Gamification is kept secondary to structured practice so engagement does not replace difficulty calibration.
Revisit
Next directions include deeper cross-linking with competition calendars (via Olympiads.co.uk), richer tutor dashboards over class cohorts, and expanded written-feedback coverage across more problem types and competition formats.