web
Primary: Korean developers (students, junior engineers, career changers) preparing for coding interviews by practising algorithm problems. They come back to the product daily or near-daily, sit in a desktop editor for focused sessions, and want to know what to solve next rather than picking problems by gut.
Secondary: hiring reviewers and recruiters who meet CodeBite as the author's backend portfolio project (README, docs/portfolio/). They evaluate it; they do not practise on it. Design decisions favour the learner.
Admins maintain problems, test cases and users through the admin dashboard.
CodeBite turns deliberate practice (Anders Ericsson, Peak) into software. Its three conditions map onto three product systems:
- Problems at the right level → the learning roadmap: 19 topics in prerequisite order. Every topic is open; the arrows show what to learn first, and each topic shows its progress.
- Immediate feedback → Run (synchronous, sample cases) and Submit (asynchronous judging via Kafka + Judge0, results pushed over SSE).
- Spaced repetition → after an Accepted submission the learner rates confidence (Again / Hard / Good / Easy), which schedules when the problem returns, Anki-style.
Success means a learner's practice time drifts toward their weak problems and topics, and they keep practising daily.
LeetCode and NeetCode give you problems and lists; CodeBite schedules your practice. It combines a prerequisite-ordered roadmap, instant judging and spaced review in one loop, with problem descriptions served in Korean and a curated set (CB50 / CB100 / CB200, 200 problems total) for Korean developers for whom Grind75/NeetCode are unknown or held back by English.
- Code workspace: three-pane layout (description · Monaco editor · test results), resizable and collapsible panels, Vim/Emacs keybindings, custom test cases, problem diagrams (trees, graphs, grids, linked lists), solve timer, per-language autosave, shortcuts (
Cmd/Ctrl + 'run,Cmd/Ctrl + Entersubmit). - Languages: Java, Python, JavaScript, C++, C.
- Daily-habit mechanics: Streaks, Repairs and Repair Progress, Review Streak (terms in
GLOSSARY.md; days are UTC perdocs/adr/0001). - GitHub Link: Accepted submissions become Solution Commits in the learner's own Solution Repository; a one-time Import backfills earlier ones.
- Surfaces: landing page, problem list, problem workspace, roadmap, reviews dashboard, profile, admin.
- Sign-in with Google or GitHub. Dark mode exists.
- Runs in production at code-bite.com on a single home mini PC.
- Korean and English are equal locales: every surface must work equally well in both, including text length, line breaking and type rendering. Problem translations are served by the backend.
- Domain vocabulary is fixed by
GLOSSARY.md(Streak, Streak Day, Missed Day, Repaired Day, Repair, GitHub Link, Solution Repository, Solution Commit, Import); UI copy uses those terms and avoids the listed alternatives. - Submission feedback latency is a product promise: measured submit-to-result is ~1.9s for Python/JavaScript, ~3.6s C++, ~8.3s Java (bounded by per-case compile/startup). UI must handle that waiting honestly.
- Existing stack: React 19 + TypeScript + Vite, Tailwind CSS v4, Radix primitives, Monaco, React Flow (roadmap), Recharts.
- Name: CodeBite.
- Logo assets:
Logo/,frontend/public/logo.svg,frontend/public/navbar_logo.svg,frontend/public/favicon.png.
- Product screenshots:
docs/screenshots/(landing, problems, editor, roadmap, reviews, profile, review-prompt). - Real measurements: submission-latency work (30s+ → 1.9s) in
README.mdanddocs/제출-지연-분석.md; load tests indocs/LOAD_TEST_RESULTS.md. - Real content: 200 curated problems, 19 roadmap topics.
- No user testimonials, user counts, or third-party endorsements exist. Do not fabricate them.
- Tell the learner what to do next. Every surface should reduce "what do I solve now?" to an obvious answer: the roadmap, due reviews, weakest tags.
- Feedback is immediate and honest. Judging state, latency and failures are shown plainly; never fake progress.
- Practice beats volume. Favour signals of retention and weak spots over raw solved counts.
- Daily return without guilt. Habit mechanics (Streaks, Repairs) reward consistency and forgive gaps; they never punish.
- Bilingual by construction. Korean and English are both first-class; nothing is designed for one and patched for the other.