Study

Study is flowss's diagram-first learning studio. Every subject in it is a connected map of concepts: you see the whole structure, work through lessons that ask you to predict, build and explain, and let spaced retrieval keep what you learned from fading. Alongside the courses sits a complete flashcard system (decks, five note types, image and diagram occlusion, FSRS-6 scheduling, a card browser with a search language), a graded assessment, a modelling canvas, an exam planner called Scholar, and the flowss Study Agent, which tutors, checks your answers and drafts models from your notes. This page is the complete reference for all of it. Open Study at /study.

At a glance

Who can use Study

Study is one of the seven studios that open on Starter and above; the Studio and Sketch are the only studios on Free. If your plan cannot open Study, visiting /study takes you to the pricing page with Study marked as locked. You must be signed in; a signed-out visit goes to the sign-in page and returns you to Study (with any link details intact) afterwards.

Everything in Study that does not call a language model works without AI: the courses, lessons (except the AI marking of a teach-back, which falls back to a local check), practice, assessment, flashcards, occlusion, the card browser, the canvas and every tracking screen. The features that need AI are the Tutor, Verify and Generate tabs of the flowss Study Agent, New course (AI), Add a unit on your own course, Generate with AI decks and Add cards with AI. They work on Starter with your own AI key (Bring your own AI key) or on Plus and above with hosted AI points (AI points and limits).

Your first visit: the welcome tour

The first time you open Study a five-step welcome overlay, Welcome to Study, appears. A progress bar across the top shows where you are, and Skip (top right, or the bottom-left button on the first step) closes it for good. Skipping still counts as finishing, so it will not reappear; you can replay it any time from Preferences → Replay the welcome tour or from ⌘K → Replay the introduction.

Start learning saves your choices and opens the course you picked on its Learning Map. Every choice stays adjustable from Preferences.

The workspace

Study is laid out like every other flowss studio: a header bar, a collapsible learning tools pane on the left, the workspace in the centre and the collapsible flowss Study Agent pane on the right.

The header

The Flow dock

Like every flowss studio, Study carries the small floating Flow pill (drag it anywhere; double-click to snap it home). Select it or press ⌘J to open the Flow dock: your session's phase (Warming up, In flow, Deep work, Scattered or Idle), with a link to the full reading in the Flow console; Continue, your latest work in the other studios; Journey, your recent hand-offs between studios; and Feedback to the flowss team. The dock's Vault keeps checkpoints of canvas studios such as Weave; your courses and decks live in Study itself and in your account's sync, not in the Vault. See Weave for every section of the dock, and The Flow layer.

Panes on wide and narrow screens

The left pane is 192 pixels wide and docks beside the workspace from 768 pixels of window width upwards; below that it slides over the workspace (slightly wider, at 224 pixels). The flowss Study Agent pane docks from 1,024 pixels upwards (resizable between 280 and 560 pixels by dragging its left edge, default 330, and never more than 30% of the window; double-click the edge to reset it) and slides over the workspace on narrower windows. When a pane is covering your work, tap outside it or press Escape to close it. On a phone, choosing a screen in the left pane closes the pane so you can see the screen you asked for.

The first time you visit, each pane opens only if the window has room for it, and the flowss Study Agent pane opens only once a course is open. After that, Study remembers your own choice for each pane.

Study remembers where you were

Study reopens on the screen and course you were last on, including which tab of Progress and Review plan you had open. Two screens are deliberately not restored: New course (AI) and Edit this course reopen on Home. If the course you were in is no longer on this device, the workspace says "No course is open" and offers Choose a course and Create one with AI.

The learning tools pane

The left pane lists every screen in five groups: Learn, Build, Library, Track and Teach. Every screen is also reachable by name from ⌘K, which is how you move around Study quickly; Study gives screens no single-key shortcuts.

Learning Map, Lessons, Practice, Assessment and Capstone belong to a course. Until a course is open they show a padlock; selecting one tells you "Pick a course on Home to use …" and takes you Home. The foot of the pane shows the Active course with a Switch course link, or "No course open" with Choose a course.

Home

Home opens with Study through diagrams and four cards that answer the questions you arrive with:

The Continue card is greyed out when there is nothing to continue, and the Review card when nothing is due ("Your retrieval schedule is clear."). Below the cards are your review calendar and daily goal (see Daily goal and goal streak), a Recently built row of your four latest portfolio items, and the Courses grid.

The course grid

The grid lists your own courses first (Your courses) and then Built-in courses; until you have a course of your own it is a single All courses list. A text box, Filter courses…, narrows the grid by title, tagline and promise, and the sort menu orders it by Recently worked on (the default), Mastery or Name. Create with AI opens the course creator.

Each course card shows the title (with a yours badge on your own courses), its tagline, the number of concepts, lessons completed out of the total, and your average mastery as a percentage and a bar. Select the card to open the course on its Learning Map. Hover or focus the card for its actions (on windows narrower than 1,024 pixels the buttons are always visible):

A Flashcards row closes the page with your deck count and how many cards are due; select it to go to Flashcards.

A diagram received from another studio

When you send a diagram to Study from another studio, Home shows a banner, "Diagram received from …", with up to three choices: Build a course (opens the course creator with the diagram's title as the topic and its source as notes), Make cards from it (opens the diagram card maker on that diagram; shown when the diagram came with an engine) and the cross to dismiss it. See Moving work between studios.

Courses

The built-in courses

Each course has a promise (what you will be able to do), units with an essential question and a reason it matters, lessons, typed relationships between concepts, and misconceptions with corrections. Every built-in course ends in a capstone (for example "Capstone: Model a System You Care About"); a course you generate may or may not have one. Built-in courses cannot be edited directly; use Customize (on the course card) or Fork (in Exchange) to make your own copy.

Creating a course with AI

New course (AI) (also Create with AI on Home) opens Turn anything into a diagram-first course:

1. Optionally type a Course focus, which guides the emphasis.
2. Drop your materials on the drop zone (Drop your materials here) or browse for them: PDF, TXT, Markdown, CSV, TSV, JSON, reStructuredText and TeX files, up to 12 files at a time. Files are read in your browser and only the extracted text is sent.
3. Or paste text into …or paste text.
4. Watch the meter: the generator reads up to 40,000 characters. Beyond that it says "Trimmed to the first 40,000 for generation".
5. Select Generate course. While it works it reads "Building the knowledge graph…". The new course opens on its Learning Map.

If some generated questions had no valid answer, they are left out and a toast says how many. Import a .json course (under "Already have a course file?") loads a course file you exported earlier.

Customising and editing a course

Edit this course (or the pencil on a course card) opens Edit course, marked "your copy", for one of your own courses. Everything is advisory: a box lists up to four "structural notes — advisory, not blocking" but never stops a choice, and your edits are saved a moment after you make them (the header reads "Saving…", then "Saved"). Done returns to the course; Export downloads the course as a JSON file, which Import a .json course can load again on any device.

Below the editor, Grow this course — add a unit with AI asks AI to continue the course. Type an optional focus ("Next unit's focus — optional") and select Add a unit; while it runs it reads "Continuing the course…". Concepts whose names match existing ones attach to your graph; everything else arrives as a new unit.

The relationship types Study understands are: requires, causes, influences, contains, is an example of, contrasts with, occurs before, supports, refutes, measures, implements, depends on, is part of, is analogous to, transfers to, is assessed by, and is commonly confused with.

Deleting a course

Deleting asks first, and says what goes with it: "Delete "…"? Its progress (2 completed lessons, mastery on 6 concepts) goes with it." After you confirm, a toast offers Undo for eight seconds, which restores the course and its progress.

The Learning Map

The Learning Map shows the open course three ways, chosen with the Map, Outline and Matrix tabs (use the arrow keys to move between tabs). The command strip above it also carries:

Concept states

Every concept carries a 12-dimension mastery vector (recognition, recall, structure, sequence, causality, comparison, reconstruction, explanation, application, transfer, evaluation and creation), grouped on the radar as Know, Structure, Rebuild, Explain, Apply and Extend. The map summarises it as one of five states:

Solid arrows are "requires" links; dashed arrows are every other relationship.

Map controls

When the map has focus: the arrow keys move to the nearest concept in that direction, Enter or Space opens the focused concept, + (or =) zooms in, - zooms out, 0 resets the zoom and F fits the map. Zoomed far out, detail fades and only names remain.

Outline lists units, lessons and the concepts each lesson teaches, with status-coloured chips, and an Also on the map list for concepts no lesson covers. Matrix is a grid of every concept against every other; read a cell as "row relates to column", and hover a cell for the full relationship.

The concept panel

Selecting a concept opens its panel on the right. It shows the concept's type, name and definition, its mastery radar with "mastery N% · state", and Why it matters. Three tabs show it as a Diagram (when the concept has a visual), Structured text (with up to eight of its relationships) or an Example (a typical case and, where there is one, a boundary case). Described reading reveals a text description of the visual. A Watch out: box names the common misconception and its correction, and Connections lists every relationship with its explanation. The buttons at the foot are Learn: … (or Revisit: …) for the lesson that teaches it, and Ask the flowss Study Agent.

On windows 1,280 pixels wide or more the panel docks beside the map and can be resized between 320 and 760 pixels by dragging its edge (arrow keys resize, Home and End jump to the widest and narrowest, double-click resets). Widen concept panel makes room for a bigger diagram; Reset concept panel width puts it back. On narrower screens the panel slides over the map.

Lessons

Lessons lists the course's units ("Unit 1", its title, essential question and why it matters) and their lessons. Each lesson row shows "revisit" when you have finished it, "N of M steps" with a progress bar when you have started it, or its effort range in minutes. A highlighted button at the top, Start here or Pick up where you left off, opens the first unfinished lesson.

Playing a lesson

A lesson is a sequence of steps. The strip at the top has Exit, one dot per step, the step count, and a notebook button ("Capture a note — lands in your notebook, anchored to this lesson"). Type a thought, question or analogy and select Save to notebook. If you finished some steps last time, the lesson opens at the first unfinished step and shows Picked up at step N · Start over.

Each step shows its narration in italics, then the activity. Continue moves on (and is disabled until you have answered steps that need an answer); the last step's button is Finish lesson.

Every step records evidence against the concepts it touches, and a build weighs more than a click. A teach-back is marked by AI when AI is available to you; otherwise (or during a graded assessment) it is checked locally against the step's key points and tells you which you covered and which are missing. On the reflect step the button reads Save & schedule retrieval when those concepts already have evidence, and Save my confidence when they do not yet; a confidence rating alone does not schedule a review.

If a question's answer key does not match its choices, the step says it cannot be answered as written and offers Skip this question.

The build bench

Build steps, the capstone and the assessment's final task use the same bench: a set of concept nodes to connect. Select a node, then a second node, then choose the relationship from the chips at the bottom (or cancel). Select a link's label to remove it. Check my model scores the model and names each relationship as right or wrong, lists missing links and warns if you connected a node that does not belong. Undo edge removes the last link, Start over clears the model, and Hint reveals up to five progressively stronger hints. On a lesson, Ask the flowss Study Agent sends your attempt to the flowss Study Agent pane.

Capstone

Courses with a capstone show it at the end of Lessons and in the left pane. It unlocks when every lesson is complete (or straight away with Explore freely). The capstone screen has a brief, a choice of case to model (when the course offers cases), the build bench, a Before you call it done checklist and the Rubric — how this would be judged with each criterion's weight. Submit capstone becomes available once your best model check is at least 70%, every checklist item is ticked and, where needed, a case is picked.

Practice and concept reviews

Practice starts with the course's due queue ("N concepts due for retrieval" with Start review). Below it, Free practice — pick your targets lets you choose any concepts and select Practice N concepts. Tasks are generated from the course graph and rotate between kinds, so you never memorise a single picture:

Keys: 1 to 9 choose the numbered answer; Enter goes to the next item once you have answered. The session header reads "Reviewing … · 2 of 5"; End session stops early. When the queue is done, the scheduler pushes each concept further out according to how well you knew it.

Today's review: one queue for everything

Start today's review (on Home's Review card, the Review plan banner, ⌘K, or a Scholar flashcard block) builds a single session across every course and every deck: due concepts and due flashcards, ordered by forgetting risk, mixed so consecutive items switch subject, and capped by your Review intensity (gentle 10, standard 20, intense 40 items). If nothing is due it says "Nothing is due right now — the schedule is clear."

Concept items use the same tasks as Practice (keys 1 to 9, then Enter or → to continue). Card items use the flashcard face and rating bar (see Reviewing cards). A card still in its learning steps after you rate it comes back later in the same session when its step is up. When only such cards are left, the session waits rather than ending early: it shows Next card in … with a countdown, says when the card is due (a card you are learning is shown up to 20 minutes early once nothing else is waiting), and offers Review it now, Undo last rating and Finish for now. Undo (top of the session) takes back the last card rating; concept answers cannot be undone. End session (top left) leaves at any time; everything still due waits for you.

The report at the end says "Clean sweep" (no lapses, no missed learning steps), "Session complete" or "Nothing was answered", and shows your accuracy, time reviewing, lapses (and missed learning steps), accuracy By course, and Coming up, the review load for the next seven days (today's bar includes anything still overdue). Undo last rating is still offered on the report, so a mis-tap on the last card can be taken back; Done closes it.

Review plan

Review plan starts with the same "Today's interleaved review — N items" banner and Start today's review button, then three tabs:

Flashcards

Flashcards lists every deck on the device ("N decks · N cards on the schedule"). Review N due reviews every due card across decks; New deck opens the creation panel. If you left a review part-way through today, a banner says "You were mid-review in "…"" and offers Carry on. With more than four decks, a filter (Filter decks…, searching titles, subjects and card tags) and sort chips appear: Due first, Recently updated, Name and Least mastered. Each deck card shows its card count, percentage mastered, new cards, lapses, and either "N due now", "Next due in …" or "Nothing scheduled".

Making a deck

The New deck panel has three tabs:

The deck screen

Note types

A note is what you write; it can make one or several cards. Editing a note keeps the schedules of cards that survive the edit. Before you save, the preview says which cards keep their schedules, which are new and which will go; after saving, a toast offers Undo.

The editor writes plain text or rich text. Its toolbar acts on the field you were last in: Bold (⌘B), Italic (⌘I), Code, Maths (⇧⌘M), Cloze (⇧⌘C; add ⌥ to put the deletion on the same card), Image and a diagram button for the chosen engine. Pressing a formatting button on a plain-text note switches it to rich text and says so (Cloze works in plain text). ⌘Enter adds or saves the note (the button reads Add card, Add N cards or, when editing, Save; an edited note also has Delete note, pressed twice). Add tags in the tags box, separated by spaces or commas. After adding, the next note keeps the same type, format and tags. On Windows and Linux use Ctrl for ⌘ and Alt for ⌥.

Card images can be pasted, dropped or picked. They must be PNG, JPEG, GIF or WebP and at most 5 MB; SVG is refused because an SVG file can carry script. Images are stored on this device.

Reviewing cards

Each rating button shows when the card will come back. Rating keys work only once the answer is showing, never while you are typing an answer, and never with ⌘, Ctrl or ⌥ held, so ⌘1 still switches browser tabs. On a touch screen, swipe a turned card right for Good or left for Again; the card says "Release for Good" or "Release for Again" while your finger is down. With Keyboard-first navigation on, a hint "Space flips · 1–4 rates the card" sits under the card.

A card forgotten eight times is a leech: it suspends itself and a toast says so ("… suspended — forgotten 8 times. Reword it, then unsuspend."). Suspended cards keep their history. The end screen says the deck is cleared and that missed cards will come back sooner.

Scheduling: FSRS-6 and SM-2

Every deck made since FSRS arrived runs FSRS-6, which models how well you know each card and brings it back when your chance of recall falls to the retention you set. Older decks run SM-2, which grows each card's interval by an ease factor. Deck options (from the deck's scheduler row, or ⌘K "Deck options — …" while the deck is on screen) is where you see and change this:

Changes are a draft until Save changes, and apply from each card's next review ("no due date moved"). Reset to defaults fills in the defaults for you to save. Every change can be undone from the same screen.

On an FSRS deck, the options also contain Optimise and Simulate a year (⌘K "Optimise FSRS for …"). Optimise fits the 21 parameters to this deck's reviews in the background, with a progress bar and Cancel, and shows a before-and-after table measured on cards the fit did not see. It offers to apply the result only when it predicts better than the current parameters. It needs at least 400 usable reviews across at least 40 cards and tells you how many more you need. Simulate a year shows, for every retention from 70% to 99%, the reviews and minutes a day the deck would ask for and how many cards it would leave remembered, and suggests the cheapest retention per remembered card once the review log holds at least 200 timed answers.

Import, export and daily limits

Open Import, export and daily limits on a deck:

Image occlusion

Add image occlusion opens an editor with a picture and masks drawn over it. Add the picture from a file, the clipboard (⌘V) or a drop. Tools: Select (V), Rectangle (R), Ellipse (E) and Polygon (P; click each corner, then click the first corner, double-click or press Enter to close it, Escape to abandon it). Everything has a key: Add a mask puts one in the middle, the arrows move the selection 1% (with Shift, 10%), Alt with an arrow resizes, Tab steps through masks, G groups the selection into one card, U ungroups, Delete removes, Enter jumps to the mask's label, ⌘Z and ⇧⌘Z undo and redo, and +, − and 0 zoom. Replace picture keeps every mask where it was.

Choose What each card hides: Hide all, guess one (every mask stays covered and the one being asked is marked) or Hide one, guess one (only the asked mask is covered; the rest of the picture is a clue). Add an optional header ("Label the chambers of the heart"), back extra and tags. A mask's label is its answer and is what a screen reader announces; masks without one still work, and the editor lists them. ⌘Enter saves.

Cards from a diagram

Cards from a diagram (on a deck, or ⌘K Make occlusion cards from a diagram) makes cards from a real diagram with no AI. Choose where the diagram comes from: Library (any diagram, canvas or BPMN process you have made in any studio), From … (a diagram another studio just sent), Course visual (a concept's diagram from a course) or Paste source. Study draws the diagram once and checks every label against the drawing; a label that cannot be hidden cannot be ticked and says why. Some diagrams are redrawn as Mermaid from their structure, in which case original colours and layout are not kept.

A note can hold up to 80 targets. Choose an existing deck or New deck…. A note's kind and diagram are fixed once it exists.

The card browser

Card browser (⌘K Browse cards) shows every card on the device across every deck, in a sortable table: Front, Deck, Due, Interval, Difficulty / ease, Stability, R (chance of recall now), Lapses, Reviews, Flag and Tags. Selecting a card opens its Card panel with the note type, tags, flag, scheduler, state, interval, difficulty and stability (FSRS) or ease (SM-2), chance of recall, lapses, reviews in a row, dates, average answer time, a "Stability over time" or "Interval over time" chart and its Review history, newest first.

Searching

Type a query in the search box. Quick searches are one click away: Due, Overdue, New, Again today, Leeches, Flagged and Suspended. A query that cannot be read shows an error with its position instead of an empty table, and the table shows every card until it can be read.

Comparisons accept <, <=, >, >=, = and !=. rated: reads the review history on this device, which only holds reviews from the day it began.

Acting on many cards

Select cards (Shift for ranges; ↑, ↓, Page Up, Page Down, Home and End move; Enter opens; Escape clears the selection) and the action bar offers Suspend, Unsuspend, Flag, Tags, Move (to another deck), Reschedule (a range of days from today, 0 is today), Reset to new and Delete. Every action offers Undo straight after; past reviews stay in the review history.

Find and replace (in the browser, or ⌘K Find and replace in cards) changes text across many notes: choose the field (Front, Back, or Front and back), optionally a Regular expression ($1 in the replacement is the first group) and Match case, and whether to look in the selected cards or all the search found. A live preview shows each change before Replace N, and an undo follows.

Assessment

Assessment is a graded, closed-book sitting on the open course, drawn from a question bank built from the course's lessons and its knowledge graph.

1. Choose a Length: Short (6 questions), Standard (10) or Deep (16). A thin course sits a shorter paper rather than repeating questions; the button shows the real count.
2. Tick "I'll answer these on my own, without hints or help." and select Start assessment.
3. For each question pick an answer and say How sure are you? (Guessing, Fairly sure or Certain), then Next. There is no feedback until the end.
4. Courses with a capstone end with Final task — build the model on the build bench, with no hints.
5. See report shows Passed or Keep going, items correct (and the model score), Calibration (questions you were certain about but got wrong, and ones you doubted but got right), accuracy By skill on the mastery radar, and a Review of each question.

The pass mark is 70%. On a course with a final task, the overall score weighs the questions at 70% and the model at 30%. A pass adds a machine-graded, self-attested record to your portfolio. Retake assessment draws fresh questions where the bank allows. Every attempt, passed or not, joins the attempt history on the gate screen.

While a sitting is open, the flowss Study Agent's tutor and answer checker are off ("The AI tutor and answer checker are off while a graded assessment is open — finish it to use them again."). Exit keeps your answers in this tab, and coming back resumes on the same questions. If you step away, the flowss Study Agent pane offers Discard the unfinished assessment.

Building and modelling

Canvas

Canvas is your own model of the open course (or a free model with no course open). Name it in the title box, add nodes from New node… or from + course concept…, and connect two nodes by selecting them in turn and choosing a relationship, which is drawn as an arrow with its head clear of the target node. Select a relationship to give it a justification ("Why does this relationship exist? What's your evidence?") or mark it uncertain (drawn dashed). Templates offers seven Blank scaffolds (Concept map, Process, Comparison matrix, Causal model, Framework application, Timeline, Architecture) and 27 subject templates on shelves for Biology, Chemistry, Physics, Mathematics, History, Economics, Computer science, Languages, Geography and Study skills (for example "Animal cell — organelles", "Mitosis — the cell cycle", "Electric circuits — series vs parallel", "Proof by induction — sum of 1 to n", "Supply & demand — market equilibrium", "Graph traversal — BFS vs DFS" and "Cornell notes — the 5 R's"). Applying a template to a canvas that already has nodes asks first. Select a node and press Delete or Backspace to remove it.

The flowss Study Agent's Turn this into a diagram loads a draft onto the Canvas. If the canvas already holds your model it asks before replacing it ("Replace your model on the Canvas? It has N nodes and N relationships…"), and the toast after offers Undo. A draft keeps at most 14 nodes.

Simulation

Simulation runs a signed causal model forward in time. Add Variable rows (name and initial value) and Link rows (source, target and a weight from −1 to +1); a positive self-link reinforces and a negative one balances. The chart re-runs as you change anything, and a read-out describes what the model did. Load a preset… offers "Reinforcing loop — growth", "Balancing loop — goal-seeking" and "Delay → overshoot & settle". Steps sets how far the run goes (5 to 500). Save adds it to your portfolio; Reset starts over.

Projects

Projects turn a course into real work: start one ("e.g. Map the burnout loop on my team"), then fill in a Brief and definition of done, milestones, linked models, evidence from your portfolio, and a decision log ("Record a decision and why you made it…"). Ship to portfolio finishes it.

Critique

Critique practises reviewing models against a course's expert model, in two modes. Model debate shows the expert model (A) and a peer's model (B) with faults injected; mark what is Wrong or reversed in B and Missing from B, set your confidence and compare. Peer review asks for One strength, One question and One justified revision, then reveals the faults the model actually had.

Personal Map

Personal Map is your own cross-course knowledge graph. Add your own idea, Pin a course concept, and Link them with a typed relationship and a reason ("Why? (the connection is the learning)"). A list names concepts "Not connected to anything yet".

Notebook

Notebook collects Note, Question, Reflection and Source entries, each optionally anchored to a concept. Questions can be marked answered, and reflection prompts ("What changed in your mental model?", "What remains uncertain?" and others) are offered. Notes you capture inside lessons land here.

Library screens

Tracking

Progress

Progress has three tabs:

Daily goal and goal streak

Set a daily goal in Preferences (⌘K Set a daily goal) as reviews a day or minutes a day, then Set goal. Every answer counts towards a reviews goal; a minutes goal counts recorded answer times, at most a minute per answer. Meeting the goal builds a goal streak, separate from the study streak in the header. You earn a freeze every 7 goal days and can bank up to 2; a freeze covers one missed day automatically, but not two in a row. Goal reminder adds a note to the bell at an hour you choose if the goal is not yet met; reminders appear only while Study is open, with no push notification and no email. Turning the goal off ends the goal streak.

Home and the Analytics tab show the review calendar: one square per study day (days run from 4 a.m. to 4 a.m.), shaded by answers. Show table switches to a table; arrow keys walk the calendar by day or week. Days before this device began keeping a review log are drawn as dashed outlines.

Calibration

Calibration plots every concept you have practised: your confidence against your evidenced mastery. Points below the diagonal are overconfident (revisit these first); points above are underconfident. Both groups are listed with a jump to the concept.

Credentials

Credentials & evidence lists credentials you have earned and those In progress, each with the evidence rule it needs: finish every lesson in a course; reach 70% mastery across an outcome's concepts; submit the capstone against the rubric; score 70% or more on the graded assessment; ship at least one project to your portfolio. Each earned credential exports a record. These are self-attested local records, not third-party verified credentials.

Preferences

Teaching preview

Teaching preview shows a course as a teacher running a cohort would see it, using a generated demo cohort that is clearly labelled as demo data ("preview · demo cohort of N"). Its tabs are Cohort (progress by outcome, mastery distribution, confidence calibration), Misconceptions (a heatmap with error clusters), Learners (one learner's model and a Suggested intervention) and Live teaching (the course map with Highlight class misconceptions and a "Predict the edge" activity). It is read-only. For real classes, see Classrooms, cohorts and assignments.

The flowss Study Agent

The right pane, flowss Study Agent ("Grounded in this course — never invents sources"), has three tabs. Ask the flowss Study Agent (on a concept, on the build bench, or from ⌘K) opens the pane on the Tutor tab with the question box focused.

These tabs need AI (your own key on Starter, or hosted points on Plus and above). While a graded assessment is open the pane is locked and says why.

Scholar: the exam planner

Scholar — exam plan (/scholar) is the deadline half of Study, on the same plan and the same data. Tell it what you are working towards (What are you working towards?): a Name ("Cell Biology midterm"), the Course, the Date and time it Starts at, and Min/day (5 to 600), optionally with Different minutes on different days under Your week (0 to 600 per day; a day set to 0 is a rest day, and without per-day minutes you tap a day to strike it out as a rest day). You can also narrow the scope to the units or concepts the exam covers; leaving it empty means the whole course. If you have no course yet, Scholar says so ("No assessment yet") and sends you to Study to build one. Build the plan plans backwards through the prerequisite graph, schedules the spaced reviews that make the material last until the day, and forecasts your readiness on the date, naming the concepts likely to cost you marks and why. Each plan block (Learn, Review, Rehearse) opens the right place in Study: the lesson, a review of that concept, the practice bench or today's flashcard session. Start in Study (or Carry on in Study) takes you to today's next block; when nothing is left it says "Today's plan is done." Edit date, scope and hours reopens the form as Edit this assessment, with Save changes. Scholar plans stay on the device where you made them; they are not part of Study's cloud sync.

Bringing work into Study

Your data: storage, sync, export and import

Study keeps your courses, progress, decks, review history, card pictures and preferences on your device. When you are signed in, cloud sync copies decks and cards, review history, course progress (including your own courses, canvases, notebook, projects and personal map) and Study preferences to your account and brings in your other devices' changes; reviews taken offline on two devices are merged, not chosen between. Not synced: pictures on cards (they stay on the device where you added them), Scholar exam plans and the goal streak's record of past goals. Pane widths stay with each screen.

Progress → Your data has:

If this device's Study data was last synced with a different account, sync refuses to merge it into yours and explains why. The card then offers to replace this device's Study data with your account's; it first says exactly how much would be removed and when it was last synced, so you can export a copy first. Leaving the card alone keeps the device's data unsynced. See Cloud sync, offline work and devices.

Keyboard shortcuts

On Windows and Linux, use Ctrl for ⌘.

Study's ⌘K commands include: Search concepts across courses, Start today's review, Browse cards, Find and replace in cards, Make occlusion cards from a diagram, Review statistics, Set a daily goal, Review … now, Drill … as flashcards, Switch course, Ask the flowss Study Agent, the pane toggles (Show the flowss Study Agent / Hide the flowss Study Agent), Explore freely (ignore prerequisites), Replay the introduction, Sync Study now, Cloud sync status, and, with a deck open, Deck options — … and Optimise FSRS for …. See Keyboard shortcuts.

Tips

• Start every session from Home's Review card: one interleaved queue across courses and decks is more effective, and quicker, than reviewing each course separately.
• Use Drill as flashcards on a course to get a deck with no AI; rating those cards also moves the course's concept mastery.
• On the Canvas, write a justification on every relationship and mark guesses as uncertain: the lint and the flowss Study Agent's verifier both read them, and an honest model gets more useful feedback.
• Use Study this on a Library document (in the card's "…" menu) to turn a diagram you already made into a course with no copying.
• If one card keeps failing, it usually holds two facts. Reword it, then unsuspend it, or split it into a Cloze note.
• Run Optimise on an FSRS deck once it has a few hundred reviews; until then the published defaults are a good fit.
• Set a modest daily goal and let freezes absorb the occasional missed day.
• Turn on Explore freely when revising a subject you already know; turn it off again for a new subject so the map guides your order.

Limits and known constraints

• Study needs Starter or above; AI features also need your own key (Starter) or hosted AI points (Plus and above).
• The course creator reads up to 40,000 characters; deck generation reads up to 20,000. Both accept up to 12 files at a time. Scanned PDFs without selectable text cannot be read.
• AI-generated decks have 4 to 30 cards per request.
• Card images: PNG, JPEG, GIF or WebP up to 5 MB; no SVG. Card images do not sync between devices.
• Learning and relearning steps: up to 10 steps, each from 1 minute to 1 day. Maximum interval: 3,650 days. Daily caps: up to 999 new and 999 review cards a day per deck.
• FSRS optimisation needs at least 400 usable reviews over 40 cards; the retention suggestion needs 200 timed answers.
• Diagram cards: up to 80 targets per note and 1 to 3 connections per complete-the-diagram card. Diagrams redrawn as Mermaid lose their original colours and layout.
• The review history only contains reviews from the day this device began logging, so rated: searches and older calendar days cannot show earlier reviews.
• Study sync sends at most 4 MB of decks, cards and progress. Above that, decks, cards and progress are not synced (reviews still are) until you bring it under the limit.
• Teaching preview uses demo data; Credentials and passed assessments are self-attested records, not third-party verification.
• Interface translations cover only a few labels; the studio is otherwise in English whichever language you pick.
• The Simulation runs at most 500 steps.
• Study has no live co-editing session; it is a single-person studio.

Troubleshooting

Related pages

• What flowss is
• Core concepts and glossary
• Keyboard shortcuts
• Your library
• Moving work between studios
• Cloud sync, offline work and devices
• Classrooms, cohorts and assignments
• Learn tracks
• AI in flowss
• AI points and limits
• Bring your own AI key
• Plans and what they include
• Privacy and your data
• The work model
• Weave