Reference
Wiki
Exact definitions for the terms and numbers Sapling puts on screen.
Knowledge graph
One node per concept in a course, joined by an edge when the two are related. Each course anchors its own cluster, so a semester reads as a handful of constellations rather than one tangle.
NodeOne concept in one course, carrying a mastery score from 0.00 to 1.00. Every concept you touch anywhere in Sapling resolves to a node.
Course rootThe labelled hub at the centre of a course’s cluster. Every concept in the course wires back to it.
EdgeA link between two concepts, drawn at a thickness that follows its strength. Stored with a type — prerequisite, builds on, part of, or related — though the tree draws all four the same way today.
Dedup“Linear Regression”, “linear regression” and a double-spaced copy all resolve to one node, so re-running an extract never splits a concept in two.
ScopeYour graph is yours, and it keys to the course rather than the section. Retake a course and your mastery carries over instead of resetting.
Mastery tiers
Every node carries a mastery score from 0.00 to 1.00, drawn as a ring around it. Four tiers are derived from that score, and they are what the tier filters, the dashboard counts, and the quiz generator all read.
Mastered0.75 – 1.00The top tier. Achievements that count mastered concepts count these.
Learning0.45 – 0.74Partly there. Still eligible for the recommendations on your dashboard.
Struggling0.10 – 0.44Repeated misses. Quiz generation leans hardest on this tier.
Unexplored0.00 – 0.09Planted but never demonstrated. Every new concept lands here at 0.00.
How mastery moves
Mastery moves on demonstrated understanding, not time spent — which means several things you might expect to move it do not. Every change is appended to your graph with a reason attached.
after = clamp(before + correct × 0.03 − wrong × 0.02, 0, 1)
That is one quiz. Around seventeen correct answers in a row take a concept from nothing to mastered — roughly three or four full quizzes.
Quiz answer+0.03 / −0.02Per question, right or wrong. Mastery is earned faster than it is lost, so a rough quiz dents progress without erasing it.
Tutor turn−0.10 … +0.30The tutor proposes a change per concept and the range is enforced, so no conversation can hand you a mastered node.
New concept0.00Whatever planted it — a document, a note, a tutor session — a concept arrives unexplored. Extracting seeds your tree, it does not grade it.
Flashcard ratingno changeRating a card records the rating. It does not write to the concept the card tests.
Assignment scoreno changeYour gradebook tracks the grade on your transcript. It runs alongside your graph rather than into it.
Progress stats
The numbers around the app that describe your movement rather than your position.
Mastery eventEvery change is appended as a timestamped delta with a reason, never written over a running total. Your graph keeps the whole history of how a concept moved.
StreakConsecutive UTC calendar days with study activity. Any activity counts once; miss a day and it restarts at 1.
Learning velocityMastery gained per day across the last 14 days, counting gains only. A node with no recent events reads 0.
Times studiedHow many study actions have touched a concept. It ticks up whether the action moved mastery or not.
Adaptive quizzes
A quiz is written fresh each time from your graph, weighted toward the concepts you have mastered least. These are the values the quiz screen builds its selectors from, so it can never offer one the server refuses.
Length3, 5, or 10Ten is a hard ceiling, not a preference: the model rejects a longer structured request outright.
DifficultyEasy · Medium · Hard · AdaptiveAdaptive hands the choice to the generator, which picks each question’s level and reports back the mix it settled on.
Question typeMultiple choiceThe only type today. Grading is an exact match against the correct option, never an interpretation.
Generation limit8 per 5 minutesSized for a person comparing difficulties or retaking a concept, not for a loop.
Generation timeout90 secondsPast this you are told to try again rather than left on a spinner.
Abandoned after24 hoursA quiz is one sitting. An attempt left open longer stops being offered as resumable.
Tutor modes
Three ways to work the same concept. You pick one per conversation, and it changes the tutor's stance rather than its knowledge — all three read your graph and the documents you uploaded.
SocraticLeads with the question that gets you to the answer yourself. The default, and the mode every Learn this and Explain this link opens in.
ExpositoryExplains the concept directly, grounded in the material you uploaded for that course.
TeachBackYou do the explaining and it plays the student, naming the step you skipped.
Uploads
What Sapling accepts, and what it does to read it.
File typesPDF · DOCX · PPTXAnything else is refused at upload. Photos and images are not accepted here.
Size limit100 MBPer file. A larger one is rejected before any processing starts.
Text extractionDocling, then TesseractA layout-aware pass first, falling back to plain OCR. A clean digital PDF reads better than a scan, and a scan better than handwriting.
RetriesProcessed onceAn upload that gets retried — a flaky connection, a second tap — is recognised as the same request and not run through the pipeline again.
Ingestion
What happens to a file after you drop it in. These five steps stream to your screen as they run, so you can always see why a concept or a date exists.
01Extract textThe file is read page by page, with OCR where the page is an image rather than text.
02ClassifySapling works out what it is holding — a syllabus, a lecture deck, a reading — and writes a short summary.
03Extract conceptsEvery concept named in the material is pulled out, along with the notes backing it.
04Update graphThose concepts are merged onto your tree as new unexplored nodes, deduplicated against what is already there.
05FinaliseThe document lands in your library with its summary and concept list attached. A syllabus also goes on to produce assignments and dates.
Notetaker
Your own writing, turned into concepts on the same tree everything else feeds.
Autosave800 msTyping settles, then the note saves. Switching notes or closing the tab flushes it immediately.
Summarise2 – 4 sentencesFaithful to what you wrote, including the questions you left yourself. A near-empty note is told it is near-empty rather than embellished.
Extract conceptsup to 15Title-Case ideas only — not assignment titles or page numbers. Each is merged onto your graph and linked back to the note.
Generate quizweakest linked conceptPicks the lowest-mastery concept linked to the note. Needs at least one link before it will run.
Flashcards
Flip a card and rate how it went. Three ratings, and what each one actually does.
ForgotKEY 1
HardKEY 2
EasyKEY 3
A rating is recorded against the card: your latest rating, the time, and one more on its review count.
There is no review schedule yet. A card you forgot does not come back sooner than one you found easy — you choose what to study.
Generating a deck reads the concepts you are weakest on. Studying one does not write back to them.
Study guide
One exam-ready review per course, assembled from what you have already uploaded.
Built fromYour document libraryThe summaries and concept notes from what you uploaded for that course. Not from your graph — a guide does not yet lean harder on your weak concepts.
ScopePer course, per examSaved once generated, so coming back to it is instant.
RegenerateOn demandRebuilds from scratch, which is what you want after adding material.
Syllabus and calendar
The one document every course hands you, turned into a working plan. One upload shows up in three places.
From a syllabusAssignments, dates, weightsYou review what was found before it saves. Due dates come through only where the syllabus states them plainly.
Re-uploadDeduplicatedAn updated syllabus will not double your assignment list.
Grading categoriesInto your gradebookThe weight buckets a syllabus names — Exams 40%, Homework 20% — seed the categories your grade is computed from.
Google CalendarImport, sync, exportOptional, and disconnectable. Sapling can also suggest study blocks around what is due.
Grade scale
Your gradebook tracks the grade that goes on your transcript. These are the letter bands a percentage maps onto by default.
A≥ 93A−≥ 90B+≥ 87B≥ 83B−≥ 80C+≥ 77C≥ 73C−≥ 70D+≥ 67D≥ 63D−≥ 60F< 60
A category scores as points earned over points possible, after any dropped lowest. Your course grade is the weighted average across categories.
The bands above are the default. You can set your own per course, because your professor’s A− might start at 92, and a curve can be applied on top.
Letters roll up to the standard 4.0 scale for a term GPA: A and A+ are 4.0, A− 3.7, B+ 3.3, down to F at 0.0.
Study rooms
A chat scoped to the people studying the same thing, next to a comparison of where you each are.
JoiningInvite or publicJoin a room you were pointed at, or browse the public ones.
MessagesReply, edit, delete, reactYours to edit or delete. The room also keeps an activity feed of what has been happening.
OverviewShared concepts onlyWhat the room sees of you is your mastery on concepts you have in common — never your notes, documents, or grades.
MatchingIn-room or school-wideSuggests partners by where their graph is strong and yours is not.
Class intelligence
The anonymised, class-wide layer. It is the least-finished thing in Sapling, so here is exactly how far it reaches.
What it isA per-class rollup of the concepts people commonly get wrong, aggregated across students with no names attached.
Where it landsQuiz generation reads it, and the misconceptions become the tempting wrong answers. That is the whole of its reach today.
What it is notThere is no screen for it. It is plumbing meant to make your quizzes sharper, not a dashboard about your class.
Onboarding
The short path from signed in to on the tree.
Sign-inGoogle, domain-gatedGated to a school email domain — bu.edu by default.
Asked onceYear, majors, learning styleLearning style is stored, but it does not shape the tutor much yet.
CoursesFrom the catalogSearched by name or code, then enrolled into the current term.
Achievements
What you earn as you go, and what it unlocks.
TriggerEach achievement watches one real number — quizzes completed, cards reviewed, concepts mastered, login streak, rooms joined. Nothing is awarded for time logged.
CosmeticsFour equippable slots: avatar frame, banner, name colour, and title. You unlock them through achievements or through a role you hold.
FeaturedYou choose which achievements and which role show on your public profile.
Your data
Sensitive columns are encrypted at rest with AES-256-GCM — a fresh random nonce for every value, and a tag that makes tampering detectable.Encrypted: your name and bio, your email, tutor and study-room messages, note titles and bodies, document summaries and extracted text, assignment points and notes, quiz questions and answers, flashcard fronts and backs.
Some things are deliberately not encrypted, and you should know which.Concept names and mastery scores, note tags, your username and cosmetics, and generated study-guide text are stored in plain form. Access to them is scoped to you rather than hidden by encryption — so keep anything sensitive out of a tag.
Your material builds your graph and your study tools. It is never sold, and never used to train outside models.Text is decrypted server-side, in flight, only to build the prompt for a study action you asked for.
Deleting a document removes it from your library, but the concepts it planted stay on your tree.That is on purpose — they are yours now, and you may have studied them since. Remove one from the tree itself, where a node can be deleted directly.
You can export everything Sapling holds on you, or delete your account outright.Both live in your profile settings.