Skip to content
I want to learn git and Linux basics
Done. Exam points aligned — questions, cards and lessons are ready. Let's drill.
QuizCardsLessonsWrong
In git, what is the first step to get a working-directory change into the repository?
A. git push
B. git add ✓
C. git commit
Working dir → staging (git add) → repository (git commit) → remote (git push). The three-area model is git's core mental model.
Question 3 / 24 · streak 5

An AI coach that drills you until you've got it

Tell your AI “I want to learn X” — the coach aligns the exam points with you, writes the questions and lessons, then drills you until it sticks. No question bank required: wrong answers get explained by exam point and reviews schedule themselves. Free and open source.

One sentence, hand it to your AI

Copy this line and send it to any AI tool — Claude Code, zcode, Cursor — it installs the coach and the site source. Tell it “I want to learn X” and you can start drilling today

Install ai-study-kit from https://aistudykit.dev/install.md

Prefer typing commands yourself? Per-tool install matrix · Want to run locally? The clone route lives there too

What happens after install

Four steps from one sentence to drilling

01 Say what you want to learn Tell your AI “I want to learn X”; got past papers or a syllabus? Hand them over — got nothing? That works too
02 Agree on the exam points The coach settles scope with you first: what's covered, how deep, how many questions — nothing starts until you nod
03 Get the full set Questions, lessons and flashcards are built around the same exam points: what the lessons teach, the quizzes test
04 Drill until it sticks Graded on submit, wrong answers explained by exam point, reviews queue up on schedule

From then on every session starts with /ask-coach: it scans your progress and picks the one thing to do now · How /ask-coach works

What it actually looks like

Screenshots captured from the live demo — not mockups. Click “Try the demo” to use it

Quiz view: a git single-choice question with correct-answer highlight and per-question explanation
Quiz: graded on submit, wrong answers tracked, every question explained
Dashboard: answered, accuracy, wrong and read stats with retry-wrong and random-drill entry points
Dashboard: progress, accuracy and drill entries computed from your answers

One toolkit, six learning artifacts

What the lessons teach, the quizzes test and the cards drill are the same knowledge points

QuizzesSingle, multi and true/false, graded on submit; multi must be fully correct
LessonsSelf-contained HTML pages with diagrams and callouts
FlashcardsCore concepts as cards, prompt on the front, details on the back
Deep-divesAI clusters wrong answers by exam point and explains each one
Spaced repetitionSM-2 schedules reviews; due cards queue up on their own
PodcastsAny study material becomes a two-host audio show for your commute

Who is this for

The repo ships a git and Linux example theme; swap in whatever you're actually studying

What you are doingDoes it fit
A developer learning a new stack, say React or K8s✅ AI distills the docs into quizzes; drill with cards
A student preparing for an exam✅ Got past papers? Import them. Don't? AI writes questions from your exam points
Getting ready for interviews✅ Say what it's for — AI writes lessons and questions, then walks you through your wrong answers
Learning anything with exam points, compliance or processes✅ If it can be Q&A, it can be learned
You just want a ready-made question bank❌ No stock questions here — but AI can generate a set from your exam points

Why not existing tools

Anki has no quiz app or deep-dives; Quizlet is closed-source SaaS and keeps your data

ToolQuizzesLessonsSRS cardsDeep-divesOpen source
Anki✗✗✓✗✓
Quizlet✓✗Partial✗✗
Notion✗Notes✗✗✗
ai-study-kit✓✓✓✓✓ MIT

Which AI CLIs are supported

Three tested by us; five more via the Agent Plugins open standard

Claude Code
zcode
Codex
Cursorstandard
Vercelstandard
GitHubstandard
AWSstandard
Microsoftstandard

Plus any other AI CLI that supports the Agent Plugins standard or a skills directory

Frequently asked questions

Free vs paid, question banks, data ownership and Anki compatibility

Is it free?

Yes. MIT-licensed open source, no accounts, no subscriptions, no telemetry.

How is it different from Anki?

Reviews run the Anki-compatible SM-2 scheduler (learning steps, graduation, lapse decay), but the loop also covers the quiz side — grading, wrong-answer deep-dives and course pages around the same exam points. It is not an Anki plugin.

Where does my data live?

In local mode progress stays in your browser; deploy the small bundled server and progress is a single progress.json file on your own machine. No third party is ever involved.

Which LLM providers are supported?

Any OpenAI-compatible API (OpenAI, Zhipu GLM, DeepSeek, Kimi, Qwen, Doubao and more), configured in your own .env. The quiz app and flashcards also work with no AI at all.

How do I deploy it?

Static frontend plus a small Hono server: pnpm build && pnpm exec pm2 start on any Node host — or run purely locally with pnpm dev.

What question types are supported?

Single-choice, multiple-choice (all-correct grading) and true/false, stored as plain JSON — no lock-in.

Can I use my own question bank?

You don't have to. Tell the built-in /ask-coach “I want to learn X”: it aligns an exam outline with you, then drafts questions, cards and lessons to match. Prefer your own? Questions are plain JSON per theme — import past papers anytime.

Can I turn my study material into a podcast?

Yes. Any material — lessons, questions, wrong-answer deep-dives — can be synthesized into a two-host audio show with a transcript. Generated with your own API key, in five languages, for passive review on your commute or while exercising.

More than a quiz tool

A study methodology distilled from real practice is built in

Syllabus defines scope→ Materials build concepts→ Quizzes validate mastery

Read the full methodology