How System Design Mastery works.

A world is one week of the course: 12 of them, from Week 00 Foundations to Week 11. A typical world has a Preview and 5 day lessons, then a boss (a timed mock interview, the diamond) and, for Week 00, Week 01, Week 02, Week 03, Week 04, Week 05, Week 06, Week 07, Week 08, Week 09, Week 10, Week 11, a trial: the week's multiple-choice questions split into parts (the T boxes). Week 00 is different: eight Foundations parts instead of days.

A lesson is split into stages (Part I, Part II… or Phase 1, 2… in a boss). Stages are how progress is saved: you resume by stage, not by scroll position.

Click a world to see its lessons with real reading times, the boss, the trial parts and the world reward: clear every lesson, the boss and every trial part of a world for +300 XP and that world's badge. On a phone, tap a world to open the same panel as a sheet.

Side quests (36 of them) are real systems designed end to end: UPI, WhatsApp, Netflix, Uber and more. They are bonus problems, open any time, worth +120 XP each, and each has a video lecture. The design sheets → link next to their title opens every side quest's design sheet, for revising it later.

The classic interview problems and low-level design live in the Prep tab, next to the company pages. Under the side quests, Beyond the course points to Prep, the Arena and War stories, plus concepts A–Z (every concept with its own page: definition, key points, the lessons that teach it, related concepts), the reading trail and the design sheets. Below that, Read offline saves the whole course or opens the offline library, and the dashed card at the bottom, The Plan, is the 11-week roadmap.

At the end of a lesson, Go deeper · reading trail shows two or three real write-ups (company engineering blogs, papers, talks, postmortems), each in three lines: what it is, the key idea and why it's here. Mark read keeps track (signed in, it follows you). The reading trail page lists them all by world, with search and company and type filters. The cards work offline in a saved copy; only the links out need a connection.

Header and course tabs. On every Learn page the site header turns into the course header: course · prep · arena · ask · shorts · stories · you (what each holds is in the tabs), a search box (⌘K: lessons, side quests, problems, low-level design and case files, or a question in plain English), your level chip and the light/dark switch. On phones the same tabs live in the dock at the bottom, with a way back to the rest of the site.

Stage rail. One card per stage with its minutes. ✓ means read, the orange outline is where you are, dashed is ahead. Optional stages (exercises, appendix, further reading, summary) sit in a separate box: they never block completion. Save checkpoint bookmarks your place (more in progress rules).

Left and right columns. The course path on the left lists this week's lessons, the boss and the trial with ticks; “On this page” on the right is the lesson's contents. On phones, tap the stage strip for the same contents as a sheet.

The chrome hides while you read. Scroll down and the header slides away; a thin stage bar stays pinned with the stage, minutes left and Save checkpoint. Scroll up a little and everything comes back.

Reading tools (bottom right on desktop, the bottom toolbar on phones): Listen reads the lesson aloud in a recorded natural voice (more in Listen); Focus hides everything but the article and goes full screen where the browser allows it (Esc to leave); Aa cycles the text size 17 → 19 → 21 px and remembers it; TERMS turns the term cards on or off (more in the concept map). Phones also get Contents (with the same term-card switch) and a shortcut to the Notebook. The orange Ask AI button above them opens the built-in AI for this lesson (more in AI in the course).

Quick checks are short multiple-choice blocks inside lessons. Pick an answer to see whether you were right and why. A right answer on the first try is worth +5 XP; get every quick check in a stage right first time for +10 more; your first fully aced block earns the “First blood” quest (+25). Retry is always allowed but only the first try pays.

Diagrams are drawn in the site's colours and follow the light/dark theme; open any of them full screen. Code blocks have Copy and collapse long listings; interview questions have a reveal card where you can mark “knew it” or “review again”.

Step-through diagrams have a ▶ Steps button: it lights one hop at a time in orange, with a caption, and you move with ◀ ▶ or the arrow keys (also full screen). Explainers are animations that walk you through an idea step by step. Every lesson and side quest opens with one that walks the whole article (its tabs are the article's parts), and smaller ones show a single idea, like a hash ring gaining a node or a circuit breaker tripping. They play when you scroll to them; pause, step, switch tabs or change the speed. With reduced motion turned on in your system they don't auto-play: you step through them yourself. Both work offline in saved lessons.

🔊 Narrate plays an explainer with its captions read aloud by your browser's voice, moving through the steps and tabs by itself, like a short video. Only one explainer talks at a time, and starting one stops Listen. Break it labs add sliders under the animation (traffic, cache TTL, replicas, timeouts…): push a number until the system falls over and watch where it breaks.

Diagram icons. Boxes that stand for a real technology carry its logo (Kafka, PostgreSQL, Redis…) and generic parts get a simple icon (a user, a phone, a queue, a load balancer). The label stays under every icon, and monochrome icons take the theme's colours in light and dark.

Practice this cards appear at the end of lessons that have a matching game, for example the Latency Ladder after Networking. They open an Arena game or a Playground board.

Pick your interview date, the days you can study and the minutes per day, and choose what to include (core lessons, bosses, trials, mock interviews, side quests), how many revision days to keep at the end and whether to leave a catch-up day each week. You get a day-by-day plan from your real progress. Fall behind or race ahead and the plan moves; the page says by how much and offers to keep the adjusted plan.

Download .ics puts every session in Apple Calendar or Outlook (re-download after a change and the sessions move); Add to Google Calendar explains the import and has a link for each day. The plan's settings are saved with your XP, so signed in it follows you. The progress page shows today's session next to your weekly goal.

There are 184 lectures (86 course lessons, 8 recorded mock interviews, 36 side quests, 20 interview problems, 22 low-level design, 12 case files), about 55 hours in all. Diagrams build up while the narration explains them, requests travel through them, and the numbers, timelines and phone screens appear as they are mentioned. The lectures link back to the earlier lessons they build on (the ↩ recall chips), and when one ends the next one in the same track starts by itself.

Watching is free with an account. Signed out, a lecture plays its first minute, then asks you to sign in. Share in the player gives you a link that can start at the moment you are on; anyone who opens a link you shared watches all of it, no account needed. The 👍 next to Save offline upvotes a lecture (one vote each, signed in), and the header and the playlist show how many people have watched it.

Quiz checkpoints pause the lecture and ask about what was just shown; a right answer earns +10 XP (once per question), and the chapter list marks each one ✓ or ✕.

✦ Ask about this moment (A) pauses the lecture and opens a panel over the right of the stage (a sheet on phones). Under each chapter, learners often ask here has the common questions already answered, free to read. Signed in, you can ask your own or tap a quick question (explain it simpler, a real example, why not the naive way, what an interviewer would ask, in Hinglish). The AI answers with the paused moment in front of it, can light up part of the diagram and can offer a chip that jumps to another chapter. Closing the panel resumes the lecture if it was playing.

The playlist shows how much you have watched overall and in each section, continue watching for lectures you started (or what comes next), and a ✓ watched mark once you reach 90% of a lecture. A lecture resumes where you stopped; signed in, that follows you to other devices. The Progress page adds your watch time. Every lecture carries a difficulty tag:

Listen sits in a lesson's reading tools (bottom right on a computer, the bottom toolbar on a phone), and in side quests, problems and case files. It starts at the paragraph near the top of your screen and lights up each paragraph as it reads it; skip a paragraph back or forward and change the speed in the player strip. When an article changed after its narration was recorded, your browser's own voice reads it until the new recording is out.

Listening counts as reading. While Listen plays, the time counts toward clearing the stage even if you don't touch the screen, so a lesson can clear on a walk. An explainer's Narrate and Listen never talk over each other: starting one stops the other.

The podcast page lists every episode by section, with how to follow it: in Apple Podcasts use Follow a Show by URL; in Pocket Casts, Overcast, AntennaPod or Podcast Addict, add by URL; anything else takes the RSS feed. Copy the feed from the page. Every episode links back to its animated lecture, for when you want to see the diagram.

Side quests (36 of them) are real systems designed end to end: UPI, WhatsApp, Netflix, Uber and more. They sit under the worlds on the course hub, open any time, worth +120 XP each, and each has a video lecture. The design sheets → link next to their title opens every side quest's design sheet, for revising one later in about ten minutes. They make good weekend practice once a world or two is done.

A case file is closed when you have read it: the same reading rules as lessons apply to its checkpoint sections, or use “Close the case” once most of the reading time is in. Reading your first case file to the end is a quest (+50 XP) and earns the On-call badge. A case can also end with a debrief question (the Tatkal case does): make the right call for +50 XP, one answer per case.

13 case files end with an interactive activity: a simulation or spot-the-bug drill in a few steps. Score 70% to solve it for +25 XP; rerun it as often as you like to raise your best. Solve all of them for the Case cracker quest (+150 XP) and badge.

Every case file opens with an explainer that plays the whole story (with Narrate), and a new case file comes out every Monday. Follow the RSS feed to hear about each one.

Every case file also has a lecture: the story told on screen, with the timeline, what users saw and the mechanism drawn out. The “watch it first” card sits under the cover.

The built-in AI answers from the case file you are reading. Open it three ways: Ask AI in the header next to Save offline, the Ask AI about this case button that stays in the corner while you read, or select any passage and press Ask here in the toolbar, which quotes that passage in the question. Pick a style (just answer, explain simply, give an example, say it in an interview, in Hinglish) and ask. On a computer the panel docks on the right and the page makes room for it; on a phone it opens as a sheet. Asking needs an account; credits work as described in AI in the course.

The same selection toolbar also has Ask Claude and Ask ChatGPT (a new tab with the passage and the case as context), Copy prompt and Copy link, which reopens the case file at that passage.

Real stories are short and plain: what happened, the numbers, and why it matters, written for anyone (you don't need the course to follow them). Every one lists its sources, the company's own engineering blog or postmortem first; if we couldn't check a story, it isn't there, even a famous one. The page picks a story of the day (the same for everyone, changing at midnight IST); filter the rest by theme: scale surprise, outage, live migration, money, people, strange but true. Stories you finish are ticked in this browser.

War rooms drop you into an incident as the on-call engineer. The channel fills up as the night goes on: the page, charts, logs, and messages from the incident commander, the experts and, later, the CTO. At four to six points you make the call. A wrong move costs minutes and users, and someone tells you why; then you choose again. Escalation runs the way it does in real teams: on-call acks, an incident commander takes charge, experts are pulled in, leadership is kept informed and makes the business calls.

At the end you get your run (wrong calls, minutes lost, users hurt), the blameless postmortem, and the real incident it is modelled on, with a link to the real postmortem. The company and people are fictional; the failure chain follows the real report. Clearing a war room is worth +60 XP once; replay it as often as you like, your best run is kept in this browser. Use fast or skip to the next call if you read quicker than the messages arrive.

Some interviews have an “implement this component” round. A code lab is one of those, taken from a lesson: a token bucket, an LRU cache, a consistent hashing ring, an outbox relay and more. Read the brief, write Python 3.14 in the editor and press Run tests (⌘/Ctrl + Enter). Python runs on your device, in the browser: no install and no server. The first run downloads it once (~6.5 MB).

Each lab has example tests you can read and hidden tests for the edge cases, shown after a run (they ship with the page, so they are practice, not an exam). Hints are there when you're stuck; the solution unlocks after 3 runs, or earlier if you insist. Pass every test to clear the lab: +60, +90 or +120 XP by difficulty, once, even after peeking. Your code autosaves in this browser. Lessons with a lab end with a Try it in code card.

Offline: save a lab for offline use and Python is stored with it (about ~13.5 MB, once for every lab), so a saved lab runs its tests with no connection. The AI hints and the code review need one.

Signed-in learners are ranked every week by leaderboard points. Everyone starts in the L1 Intern league. When the week ends (Monday 00:00 IST), the top 20% of each league's active players (at least one, and only with 60+ points) move up a league; a week with zero points drops you one league. Reaching the top 10 of the whole board in a week earns the Top 10 quest (+100 XP).

Arena games are 60–90 second rounds: Latency Ladder, Pick the Store, Fix the Outage (+40 XP each on the first clear). The Playground boards (Cache or Miss, Hash Ring, Trip the Breaker, Back of the Envelope, Race Condition, Incident Commander) are longer simulations (+60 XP each). Every card lists the lessons that teach it, and best scores stay in your browser.

Where cards come from: every question you get wrong (or skip) in a trial, right answers you flagged as a guess, and highlights you turned into cards in the Notebook. You never have to make cards by hand.

Boxes (spaced repetition). Every card sits in one of five boxes. A new card starts in Box 1. When it comes up you try to recall the answer, tap Show answer, and grade yourself:

A revision deck boils a lesson down to about 5 minutes of cards: key ideas, numbers to remember, a trade-off table, the classic mistakes and if the interviewer asks (the answer stays hidden until you show it). Open one lesson's deck or Revise this world for all of them in a row. The clock at the top counts down 5 minutes a lesson; it keeps going past zero, it never stops you.

Swipe, use the arrows or the buttons. Mark each card Got it or Again: cards marked Again join your daily standup. Finish every lesson deck of a world (in one go or one at a time) for +40 XP, once per world. Keys: ← → flip, Space next or show the answer, G got it, A again, S shuffle, Esc exit.

A cheat sheet is the same material for a whole world on one page. It sizes itself to print on a single A4 or Letter sheet; Download PDF opens your browser's print dialog (choose “Save as PDF”), and Share sends the link. Day lessons, Foundations parts and bosses end with a Revise … in 5 min card and a link to their world's sheet.

Everything at once: every cheat sheet as one PDF, one world per page, and an Anki export of all the revision cards (in Anki, File → Import: one deck, tagged by world and lesson) to keep revising in your own spaced-repetition app.

On the design canvas, pick a classic interview problem, a week's capstone or your own idea. The board is the same one the mock interviews use: building blocks with icons drop in with a click, Tech logos adds any technology by name, and an arrow drawn from one block to another stays attached when you move them. Write a few lines of notes under it, the way you'd say the design out loud.

Review my design reads the drawing (the blocks, the arrows and their labels) and your notes against the problem's brief, the way a staff engineer would: a verdict, the gaps that would sink it worst first, the fix for each, and the questions you'd be asked next. Fix the worst gap, then review again. Your designs save to your account as you draw, so the canvas needs you to sign in.

Trials are a week's multiple-choice questions, split into parts of about 15–30. Each part tests one lesson and opens when you finish that lesson (a “Locked” part still has Start anyway). Questions come one per screen, untimed: choose, then Lock in to see the answer and explanation. Your picks autosave, so a closed tab comes back to the same question.

Pass mark: 80%, rounded up (12 of 15, 16 of 20, 24 of 30). The first pass of each part is worth +100 XP. Fail and the part rests for 10 minutes, then comes back with a new option order; retakes only raise your best score. Everything you missed and every flagged answer goes to your review queue. Practice all puts every question of the week on one untimed page with no XP; its answers are kept apart from your attempts.

Interview mode: each phase has the capstone's own time budget (the clock runs over budget, it never cuts you off). You write your answer in a scratchpad first; the model answer stays hidden until you reveal it. At the end you grade yourself against the capstone's checklist: 70% defeats the boss for +200 XP, once. Read mode is untimed with the answer open and gives no boss XP. Reading the capstone to the end still completes it as a lesson (+60 XP) and marks the boss node done on the tree.

One open system design problem a week. Read the brief and its constraints, write your design in markdown with Mermaid diagrams (it autosaves and stays private while you write), then vote on the community's designs once the gallery opens. Winners are featured on the page, on the course hub and on their profile. Submitting is worth +40 XP, a featured design +100 and the winner +250. Reading is open to everyone; submitting and voting need an account and a connection. Until the first challenge opens, the page says “coming soon”.

There are 8 mock interviews, numbered in course order, one after World 01 (Paisly), World 02 (Billbolt), World 04 (Cartwala), World 05 (Wanderstay), World 07 (Sunwaai), World 08 (Kaamkaaj), World 09 (Switchyard) and World 10 (Tillsy). Each covers everything taught up to that point, not just the last week. They appear on the skill tree as their own rows between worlds (the square node with an orange corner), in the “After this world” card of a world's panel, on the Prep page and all together at /mock.

Each mock opens on its lobby: the recruiter's note, the interviewer and their style, the fictional company, and on a wide screen a cover that sketches the problem like a whiteboard. Below that, watch it first plays the same round recorded: a model interview in two voices (see video lectures). Then the phases and their budgets, and the level you are interviewing for: Junior, SDE, Senior or Staff.

The Whiteboard tab is a real diagram canvas. Building blocks with icons (client, users, CDN, load balancer, API gateway and service, worker, database, cache, queue, stream, object storage, search index, scheduler, monitoring) drop in with a click, in a free spot. Tech logos adds any technology by name (Redis, Kafka, PostgreSQL…), and + Custom block makes your own: a name, a shape, a colour and an icon, kept in my blocks for next time. Draw an arrow from one block to another and it stays attached when you move them. Full screen gives it the whole window; the menu exports a PNG. Notes under the board are shared across phases.

When the call ends you see what happened (time used, phases over budget, curveballs answered, hints, pauses, words written), any red flags, your whiteboard as a picture with its blocks and arrows written out, then each phase: what you wrote, what the live coach said, the model answer behind a reveal button, and the rubric. Every rubric signal is tagged with the lowest level at which interviewers expect it, with links to the lessons that teach it. Score each one missed, partly or nailed. Signals at or below your target level are expected; the ones above it are a bonus. Until you reveal a model answer, ← Back to the call returns to the round.

Red flags from how the round went drop the verdict one band each, two at most: phase 1 under 15 words: jumped in without clarifying; any other phase under 10 words; a curveball that arrived but got under 5 words; more than 25% over the round's time; four or more hints. You also see the level your answers read like, what a hire looks like at each level for that problem, and the lessons behind the signals you missed.

Second opinions: signed in, the AI scorecard grades every phase against its rubric, quoting your words; it sits beside your own scores and doesn't change XP. “Ask Claude” or “Ask ChatGPT” opens a new tab with your answers, the curveballs, the rubric and your self-scores prefilled, asking for interviewer-style feedback and its own verdict. A round too long for a link goes through your clipboard.

Submit scorecard once every expected signal is scored (and you wrote at least 40 words in the round): +150 XP and the loop's badge, plus +100 XP the first time a verdict is Hire or better. Each is paid once; retakes go into your history and can raise your best verdict.

Practise with a real person: on peer mock interviews, post a time slot and pick one of the course's mocks and a role (candidate, interviewer or either). When another learner pairs up, you both get an email with a free video room; the interviewer runs the round from the mock's page (phases, curveballs, the rubric), then you swap. Only your first name and email go to your partner; times show in your own time zone. Posting and pairing need an account.

The Prep page opens with the video lectures (the whole playlist, a chip per section and where to start), then mock interviews, one card each: start the round, or watch it done well in its recorded version. The rest is grouped as who's hiring (company pages, the question bank), classic problems (the problem bank, low-level design), interview night (revision decks, cheat sheets, design sheets), the real loop (code labs, mock interviews, peer mocks) and tools (numbers playground, study planner). The row under the title counts what you have revised, the labs you cleared, the mocks you finished and your study plan. All of it works as a guest. Saved for offline use, each card says whether its page is saved on this device; the rest show needs a connection.

Each company page says what the company builds, the topics its questions lean on, and two lists. Reported by candidates: questions people wrote up after interviewing there, each with a link to that write-up. Likely, from what they build: our pick of questions that fit its products, clearly labelled as ours. Every question links the lessons that teach it, and the page ends with a study path through them. No company publishes its questions; we only call a question reported when a write-up we could read says so.

The question bank merges every company's questions, most-reported first: filter by type (system design, low-level design, concept), by company or to reported ones only. A filled dot next to a company means reported with a source; an outline means likely.

The problem bank (the classic questions: URL shortener, rate limiter, chat, news feed…) and low-level design (machine coding: parking lot, LRU cache, Splitwise…) are listed on the Prep page too. Each opens with a mission: the brief and the numbers you design for.

A design sheet is a side quest's whole design on a few pages: the numbers at a glance, the architecture with real tech icons, the key flows as sequence diagrams, the data model as a class diagram, the core entity's state machine, the decisions (chose, over, because), what breaks and how it's contained, and the lines to say in the interview. Read the quest once; revise it from the sheet in about ten minutes.

Download PDF prints it in the light theme with every diagram drawn, no diagram split across pages. Each side quest ends with an open the sheet card, and the design sheets page lists them all, with how many diagrams each holds.

Type an estimate the way you would say it in an interview, like 50k QPS × 2 KB × 30 days, and get the answer with the working: each step, the units, an “interview shortcut” version and the bandwidth or storage it implies. Write one step per line and name results (peak = 3 × avg); presets cover the classic questions, and copy link shares your sums.

Below it are the latency, throughput and size numbers the course teaches, each linked to its lesson (work it out sends one to the calculator), and a drill: Week 00's trial has an estimation part. Data units are decimal, as in the lessons: 1 KB = 1,000 bytes.

The pen button (N on a keyboard) opens the notebook beside whatever you're reading: a lesson, a side quest, a lecture. On a laptop or a tablet held sideways it docks at the side; on a tablet held upright the lesson stays on top and your page sits below, with a handle to resize the split. The panel keeps your recent pages one tap away, and the notebook lists every page, newest first.

Writing: pen, highlighter, eraser and lasso, with real pressure on an active pen. Scribble over strokes to erase them, hold the pen still for a quick menu (erase, recolour, thicker, thinner, turn into text), tap with two fingers to undo. Pages come ruled, grid, Cornell or interview paper.

Clip: select a line in the lesson and tap Clip; it lands on your page with a link back to the exact spot. Snip: for a diagram, code or a table, tap Snip and drag a box; it pastes as a picture, and you can ask about it (“why does locking hurt at high traffic?”). The answer comes from the lesson you're on and saves to your page. Lecture pages snip a paused frame the same way.

Turn into text reads your handwriting and writes it out as clean, searchable notes; flip between the ink and the text any time. Check against lesson compares your notes with the lesson and flags what's wrong or missing, quoting the section that says otherwise. Make review cards turns the notes into cards; untick any you don't want and the rest join your daily standup. Fixed a slip by hand? Update the text rereads the page.

Annotate lets you write on the lesson itself: circle, underline, scribble in the margin. Your marks stay on the words, even when the text size or the screen width changes. Or select a passage and write a handwritten note that stays pinned to it.

Every week of the course and every side quest has a written paper: questions built from its checkpoints, drills and design prompts. Sit it untimed (practice) or against the clock (exam) and answer each question by hand on a sheet. The AI marks every point of the question's marking scheme and shows the evidence it found on your sheet; then read the model answer and send what you missed to review. A paper's card also sits next to the week's trial.

Highlight it in one of four colours, each with a meaning: Key idea (orange), Definition, Numbers (grey) and Interview line (teal). Note attaches your own words to the passage. Bookmark saves the exact passage with an optional title. These three need a free account, so they follow you everywhere; as a guest the toolbar asks you to sign in.

Ask AI opens Claude or ChatGPT in a new tab with the passage and its context (lesson, stage, the paragraphs around it) already written as a question; Copy prompt gives you the same text for any other assistant. Copy link makes a link that opens the page scrolled to that exact line (on phones, Link also offers your phone's Share sheet). The # next to any heading copies a link to that section.

Keyboard: select text (or Shift + arrows), then H highlight, N note, B bookmark, A Ask AI, L copy link. Shift+F10 opens the toolbar.

The Notebook has three tabs: Highlights, Notes and Bookmarks. Search across all of them, filter by week or colour, edit or delete (with undo). Open at … takes you to the stage you saved it from. + Review card turns a highlight into a flashcard with its key phrase blanked out, which joins your daily standup. Export Markdown downloads everything.

Open it from the ask tab, or the orange Ask button on any course page (there it knows the page you're on). Type, or tap a quick action: Quiz me, Start a mock, Make a plan, What next? Before it answers it can look things up: your progress, the course's lessons and lectures, the week plans. You see each step as it happens.

Sources. Answers are written from the course: each claim carries a number, and the sources (the lecture, the lesson section, the questions learners ask) sit beside the answer on a computer and under it on a phone; tap one to open it at that spot. When the course doesn't cover something, the tutor says so first, then gives a general answer marked as general.

Plans and paths. Ask for a plan and it reads your progress and the course's own week plans, then lays the plan out as a learning path; Save path keeps it under Saved paths as a checklist that ticks itself off as you read.

Quizzes. “Quiz me” picks a checkpoint from what you've been reading; answer in your own words and it grades you against the course's reference answer: what you got, what you missed.

Memory. When a chat ends, a small model writes a few short notes: your goals, what you studied, what you found hard. The next chat starts from them. Read them, switch memory off or clear it on the memory page (you → memory). Past chats are in the history list.

AI credits. Each signed-in learner gets a daily allowance (shown in the AI panels and on the Progress page), refilled at midnight UTC. A lesson question uses about 10–20 credits, a mock interview turn about 10. The course chat (the Ask tab and the Ask button) doesn't use credits: it has its own daily messages, shown under the chat box. When many people use AI on the same day the allowance shrinks a little so it lasts for everyone; if it runs out for the day, the Ask Claude / ChatGPT links and the free questions still work.

Use your own key. On the Progress page you can add your own Anthropic API key. It stays in your browser, is sent only with AI requests and is never stored on our side; your Anthropic account is billed and the daily limit doesn't apply.

What's sent. What you type into an AI feature (a question, your code, your answers) goes to Anthropic's AI to answer it. It isn't used to train models. We keep the conversation text for 30 days to stop abuse, then it's deleted. AI needs an account.

Overview: level and XP with what the next level unlocks, course % (lessons, bosses and trial parts done out of all of them), your current streak (days with a cleared stage, a lesson, a trial part or a review), reading time, lessons done and a bar per week. On the right: resume where you left off, the next trial part to take and today's standup.

Your data. Download my data saves one JSON file with everything your account holds (signed in) plus this browser's copy. Signed in, an Import card offers to move progress you made as a guest in this browser into your account. Reset clears exactly what you choose, in three steps: what to reset, which data, confirm.

The concept map lists every concept the course teaches, grouped by world and coloured by how well you know it, from your quick checks, trial answers and review cards (nothing extra to do): not started, seen, shaky, solid and mastered. Tap a concept for a two-minute refresher: the definition, the key points, your evidence, the lesson that teaches it and a way to practise it.

Term cards: in lessons, course terms have a dotted underline. Hover (or tap) one for its one-line definition, where it's taught and related terms. Don't want them? TERMS in the reading dock, the switch in the phone Contents sheet or the one on the concept map turns them off for this browser.

Quests are one-time goals with an XP reward: the site's main quests (open the terminal, survive Tatkal Rush, read a case file to the end…) live in the quest log; course quests (First blood, daily streaks, Top 10, each game) are awarded as you play. Badges mark bigger milestones: one per cleared world plus site badges like First blood, Daily streak 7, Top 10, Arena regular and Case cracker. Hover a badge to see how to earn it.

Turn it on in the share section of the Progress page (it needs an account). The Public page switch is off until you flip it or press Publish my page. While it's off nobody can see your progress, and switching it off again makes your link show a 404 straight away.

Handle. Your page lives at systemdesignmastery.dev/u/<handle>. A handle is 3–24 lowercase letters, numbers or single hyphens. One is suggested from your name, and whatever you type is checked as you go: taken, reserved and offensive handles are refused. Changing it later breaks links you already sent.

Name on the page: your full name, first name + initial or anonymous (just the @handle); new pages start with first name + initial. Your profile photo appears only if you tick it, your account has one and you aren't anonymous. Then pick what visitors see (all six start on). The preview on the right follows every change, and Save changes updates a live page at once.

What a friend sees when they open your link, no account needed: your name or @handle, your level on the L1 Intern → L6 Principal ladder with XP, course %, streak, lessons and worlds done, a world map, a read-only copy of your skill tree, your Arena scores, badges and recent milestones. Each part is there only if its switch is on.

The page ends with an invitation: Start the free course or Take today's daily challenge, plus “Can you beat …?” next to your scores. When you open your own page signed in, a “your page” bar adds Edit what's visible and Share; visitors never see it.

Where: the certificate page (from the Progress page or the row of links under the You tab). It shows a checklist with what's done, what's left and a link to finish each item. It needs an account and a connection: your progress is checked on the server, from your account's records, when you claim. Progress made as a guest lives in that browser, so sign in and import it from the Progress page first.

What counts. Lessons count once they're completed in your account. Trial parts count from your saved attempts: the answers are marked again on the server against the answer key, so a part counts only if those answers really reach 80%. If an older pass doesn't (the questions changed since), the checklist says so: retake that part. Capstones count once you've completed reading them. Bosses count when you win them in boss mode. A mock interview counts when its scorecard (you score yourself against the levelled rubric) comes out at Hire or Strong hire; the course certificate needs 3 different ones. When the course grows (a new world, or new lessons in a world), a certificate already issued stays valid as it is, with the numbers it was earned on; a claim made after that needs the new lessons too.

Claiming. Press Claim my certificate, check the name (2–80 characters, letters, spaces, apostrophes, hyphens and periods; it's signed into the certificate and can't be changed later) and choose whether it gets a public page (off unless you tick it; you can switch it on or off later, and withdraw a certificate at any time). Claiming again never makes a second one. The course certificate pays +500 XP and the Certified badge, a world certificate +150 XP.

Then: Add to LinkedIn profile opens LinkedIn with the certificate's name, issuer, date, ID and link filled in; Download PDF opens a one-page A4 landscape version, public page or not (choose “Save as PDF”); Share link opens the share sheet; Download Open Badge gives a signed Open Badges 3.0 credential for badge wallets (with the public page on, there's also a badge URL for Badgr and Canvas Credentials). The course certificate is paper-style with a QR code; world certificates are drawn as blueprints.

The offline library lists everything saved on this device, by type: the course (by world), video lectures (by playlist section), side quests, interview problems, low-level design, mock interviews, case files, war stories, revision and cheat sheets, code labs and tools. Everything saved shows it all with a search; pick a type to see what is saved out of what could be, switch between on device and all, and save or remove a whole type or section at once. Each row has its size and when it was saved, and update when a newer version is online.

Storage on this device splits the space into videos, articles and app files, and shows how much the browser lets the site use. If the browser might clear it when space runs low, ask to keep it asks it not to. Next to it, the app card says whether you are online and offers to install the course as an app.

Save offline under a lecture's title stores the page with its recorded narration, captions, music and poster, about 10–20 MB a lecture; the button shows the size before you save. A saved lesson's Save offline panel offers its lecture too, and the library's Video lectures type saves a whole playlist section (anything over 300 MB asks first). Saving the entire course leaves the lectures out, because together they are gigabytes. A saved lecture plays, shows its captions and asks its checkpoints with no connection; its progress syncs when you are back online.

Android, Chrome or Edge on a laptop: press Install the app in the offline library or the Read offline card (or the install icon in the address bar). The course opens from your home screen or dock, full screen, with your saved lessons. With no browser bar around it, the app shows its own back arrow in the header (and in the reading toolbar on a phone). On a laptop the course header doubles as the window's title bar (drag it by any empty spot); the small arrow next to the window buttons brings the usual title bar back.

iPhone and iPad: Safari has no install button, so the library shows the steps: Share → Add to Home Screen. Do this if you rely on saved lessons: Safari can clear a website's data after about a week without a visit, but an app on the home screen keeps it. Lessons saved in Safari don't carry over into the home-screen app, so save them from the app. After a save on an iPhone, the save message reminds you of this.