Home Product
ProjectsCoursesEvaluation
Technology About 中文
Log in Try demo
Technology

Everyone is making AI better at answering. We taught ours when not to.

The better an AI answers, the less a student has to think. So the engineering goes somewhere else: making a course pace itself like a teacher, training an agent to hold back, letting several agents share one project, routing support by stage and withdrawing it on time, and making every sentence of the final assessment checkable. The pedagogy is settled first; the algorithms implement it.

Five key algorithms

The pedagogy is decided first. The algorithms implement it.

Each of the five below started as a claim about teaching and only then became an implementation. They're in the order a student meets them: learn the method inside a course, carry it into a project, and have the assessment read the whole thing back at the end.

course-workflow · restraint-ladder · project-spine · stage-router · evidence-index
← swipe to pan the diagram →
her work what she leaves 01 course-workflow Human-like course workflow A course paces itself the way a teacher does: show the material, wait, ask, hand over a card, wait for the answer, move on. 02 restraint-ladder Education-enhanced AI agent One question at a time. It answers when answering helps, nudges when a nudge is enough, and hands a thinking card back when the thinking belongs to her. 03 project-spine Multi-agent assistance One agent walks every surface with an unbroken memory; work that must swallow a long text goes to an isolated sub-agent. 04 stage-router Staged AI support Framing, reading, writing, revising, review — what the AI may and may not do is different at every stage. 05 evidence-index Process evaluation The facts compute straight from the record, and every judgment has to cite a fragment that actually exists. THE THREAD THROUGH ALL FIVE The thinking card: a tool call whose executor is the student — at once the teaching move and the evidence of it
algorithms that move the learning algorithms that decide how far the AI goes reading the process back the card that runs through all five
01

Human-like course workflow

course-workflow

A course isn't a video with questions bolted on. It paces itself the way a teacher does: show the material, wait, ask about that material, hand over a card where she gets stuck, wait for her answer, then decide where to go. What she answers changes the route ahead.

The outcome of a course is capability and the deliverable is a handful of methods she can carry into real coursework — she leaves with tools in hand, not with watch time.
02

Education-enhanced AI agent

restraint-ladder

A general assistant's instinct is to complete the answer. This one asks itself something first, every turn: who should be doing the thinking right now. It answers when answering helps, nudges when a nudge is enough, and hands a thinking card back when the thinking is hers — one question at a time.

This is the line schools care about most: the AI is present throughout, and the prose and the core judgment stay hers. The system holds that line, rather than leaving it to everyone's good intentions.
03

Multi-agent assistance

project-spine

A project runs for weeks. The main agent walks framing, reading, writing and review with an unbroken memory; only work that must swallow a long text — reading a paper, assessing a whole process — goes to a context-isolated sub-agent, which comes back with conclusions only.

It carries the multi-week work — EE, IA, a research project. And that's exactly the work in which judgment becomes visible: a short task can't show you whether someone can think.
04

Staged AI support

stage-router

“Let me take a look at that” is help while she's framing a question and ghost-writing once she's drafting. So support is routed by stage: what the AI may and may not do is defined separately for each one. And support has to withdraw — a card has succeeded when it stops being summoned.

A teacher can hand the AI to a class without hesitating: the boundary is written into the system, so it doesn't depend on repeated reminders or on somebody standing over them.
05

Process evaluation

evidence-index

At the end the whole process gets read back into a report. Whatever can be counted computes straight from the record. Whatever takes judgment must point at a fragment that actually exists — click it and there's the conversation, the card, the source.

For the first time the thinking itself is something you can produce: teachers see the evidence, families read the narrative, a student sees her own trajectory — and a school finally has a second kind of proof beyond the finished work.
0
thinking cards
0
disciplinary lenses
0
observable measurement points
0
international mark schemes behind the depth axis
01 · human-like course workflow

A course runs on a teacher's pacing, not on a progress bar

A video only knows how to keep playing. Here a slice is a workflow: every step includes waiting for her, and what she does changes what comes next. A course-authoring agent writes the content; playback is fully determined — two students can end up in different places, and neither can end up somewhere broken.

a slice: play · wait · ask · summon a card · branch
← swipe to pan the diagram →
AUTHORING AGENT Course-authoring agent emits one CourseDefinition 2.0 doc THE ONE GATE validateCourseDefinition structure → references → graph. all three, or nothing RUNTIME CoursePlayer if it passes, it plays. that is the contract ONE SLICE · each slice carries a deterministic, replayable state machine answer.incorrect → retry / hint video.ended answer.correct STEP intro enterActions show(b-video) playNarration(n1) focus(b-video) STEP ask enterActions show(b-question) enable(b-question) the student's turn to judge STEP done enterActions completeSlice() the slice's behaviour is recorded 16 ACTIONS show · hide · focus · playNarration · startTimer · completeSlice … EVENTS video · narration · answer · block.completed · timer … RUNTIME SCENE · generated at play time The opening and closing are written for this student the document only carries a fallback; personalisation never bakes in COURSE SESSION · owned by the platform Progress lives apart from the course itself so the document stays deterministic, time-free, portable
written by the course agent pacing & state completion & outcome

Every step waits for her

The course moves on after she's watched, answered, filled the card in. Waiting is the one thing a video can't do — and it's the beat that matters most in teaching.

Her answer changes the route

Answer solidly and it moves on; answer vaguely and it comes at the question from another side; get stuck and a card arrives. The branches are written into the course, so the path through it is her own.

The opening and the ending are written for her

The first line shouldn't be the same for a student who just traced a source and one who just abandoned a lead. Opening and closing are generated for this student at play time, while the course itself stays put.

02 · education-enhanced AI agent

When the thinking is hers, the AI hands over a card

This is the load-bearing sentence of the whole design. The agent issues a tool call as it would any other — except this time the executor is the student: a thinking card opens in her hands and she fills it in, and what she wrote returns to the conversation to be questioned further. Here the teaching move and the evidence of it are the same object.

answer · nudge · hand the thinking back
← swipe to pan the diagram →
refeed the student's turn “This post says the Earth is 40% greener thanks to China” — with no source THE RESTRAINT LADDER · server-side First: should the thinking go back to the student? 01 · RESPOND answering is enough 02 · HINT one nudge. one question at a time 03 · SUMMON_CARD summon_card("sift", reason, nudge) auto-triggered · the student opens it CARD RUNTIME · the tool a human executes SIFT · lateral reading schema-driven: a new card is a new JSON, not new code the student fills it in: who said it, where's the original answers, hesitations and skips are all captured the AI fills in nothing. not one field STANDARD ENVELOPE · process tree grows one traceable evidence chain one structure feeds both the coach and the assessment PROVENANCE · the whole signal is in who started it the system offered it — logged as prompted nobody asked and she summoned it — logged as self-initiated
the agent's decision executed by the student teaching signal the record left behind

Four steps from a decision to a piece of evidence

The seams between the four are narrow enough to be a single function, so the pedagogy can move on its own: change when the AI should stay quiet and you touch the decision layer; add a new card and you write one card's configuration.

01 · server-side

Decide

Whether the thinking should go back to her this turn, and which card to send. The whole restraint ladder lives server-side, out of the client's reach.

02 · summon_card

Hand over

One function connects the decision to the card: summon_card(card, reason, nudge). The trigger is automatic; opening it is her call.

03 · by the student

Execute

The card opens in her hands and she fills it in. The AI completes not one field: the executor of this step is a person.

04 · tree + assessment

Feed back

What she wrote returns to the conversation to be questioned further, grows the process tree, and becomes the floor the assessment stands on.

Triggered automatically, opened by her

The act of offering a card is itself a teaching signal

The system offers a card in the right situation: logged as prompted. She summons it with nobody asking: logged as self-initiated. Same card — and that one difference is the main thing we read when asking whether the method has become hers.

03 · multi-agent assistance

One agent for the whole journey, with an unbroken memory

Framing, planning, reading, writing and review are surfaces of the same Imprint. Changing rooms moves the camera and carries the whole memory along — so in week three it still knows why she dropped that lead in week one. Sub-agents appear only when they're needed.

one continuous session · context-isolated sub-agents · the project spine
← swipe to pan the diagram →
write projection spawn spawn takeaways written back Student any surface MAIN AGENT · one continuous session Imprint system prompt (the restraint ladder) spine projection + rolling chat window active surface + its card / lens deck IN-THREAD SKILLS summon_card plan-gen outline rabbit-hole SUB-AGENT · isolated Reading agent swallows the full text, returns only takeaways SUB-AGENT · flagship, never downgraded Assessment agent reads the whole process, emits the two-axis reading THE PROJECT SPINE · durable project state The one durable state every surface reads and writes metadata proposal plan reading-list outline activity digest
the main agent reading sub-agent (isolated) assessment sub-agent (flagship) the project spine

Reading sub-agent · brief in, takeaways out

In: the project scope and why this source. Isolated on: the full text. Out: five fields — findings, credibility, key quotes, new leads, impact on the proposal. The article body never enters the main thread, so twenty sources later the conversation is still clean.

Assessment sub-agent · best model, never downgraded

The coach can run mid-tier; it only ever asks one question at a time. The assessment can't: it swallows the whole process and returns the two-axis reading plus prose that students, parents and teachers will read again and again. It's the one place we don't economise.

A project runs for weeks: the conversation can neither forget nor grow forever

The answer is to let settled things become structure: a proposal finalised, a source read, a card completed — once stable they're written into the project spine and the raw turns step out of the conversation. Each turn assembles a compact view, and detail is fetched when it's actually needed.

memory → spine → rolling window · projected per turn
← swipe to pan the diagram →
THREE DURABILITY TIERS · the lower, the more volatile MEMORY · across projects Who this student is — recurring habits, growth over time patterns and process only — never a store of her documents SPINE · this project proposal · plan · reading list · outline · activity · digest once dialogue solidifies it folds in here, and leaves the window ROLLING WINDOW · in flight the last few raw turns older ones are gone from here — and nothing is lost EACH TURN'S CONTEXT is projected the three layers are never dumped in. each turn builds a compact view on demand. so the coach never feels amnesiac
04 · staged AI support

The same help becomes ghost-writing one stage later

“Let me look at that paragraph” is help while she's still deciding on a question and ghost-writing once she's drafting. So the boundary is defined stage by stage: what the AI may do and what it may not are two separate lists at every point.

Stage What the AI does here What it never does
Framing Questions a vague interest back into an answerable one, generates search terms, flags an over-broad scope Never picks the question
Sources & reading Reads long texts in isolation and returns only takeaways; offers SIFT or CRAAP at the right moment Never decides whether a source can be used
Writing Points at a broken warrant, plays the opposition, checks whether a citation supports the sentence Never writes a word of the prose
Revising Reads the draft and comments, asking whether a change moved the argument or only the wording Never makes the edit for her
Review Reads the whole process into a report and names a next step Never writes her reflection
The point of support is to stop being needed

A card has worked when it stops being summoned

Usage on its own says nothing. We watch two curves moving together: this card's summons trending down, while the question it packaged starts turning up in the student's own prompts and in her reasons for a revision.

The stage is decided by the work itself

Where she is depends on what she has finished

Stages don't advance on a date, and she doesn't tick a box. A proposal finalised, enough sources read, a first draft in place — real events move the stage forward, and the AI's boundary moves with it.

05 · process evaluation

Facts are computed. Every judgment must cite evidence.

The report has two halves. Everything countable — turns, materials, words, comments, revisions, the timeline — computes straight from the record without calling a model at all. The half that takes judgment goes to the best model available, but it may only cite from a pre-built set of evidence, and any judgment that doesn't resolve is dropped whole after generation.

counted facts · evidence index · grounded judgment
← swipe to pan the diagram →
RECORDED DATA · everything the student left behind messages events card_instances references citations checkpoints plan_items FACT HALF · deterministic Facts that are counted, not claimed turns · materials · words · comments · revisions milestones · timeline · material list · card usage never needs the model EVIDENCE INDEX · citable candidates The only things the model may cite Every judgment must point back to a real id: a message, an event, a card, a source. unresolvable refs are dropped later 3 CONCURRENT FLAGSHIP CALLS · independent · the evaluation tier is never downgraded CALL A Prompt lens How the student asked: for output, or for a counter-argument CALL B Risk signals ghost-writing · missing sources · broken warrants · drift · tangents CALL C Two-axis rubric Cognitive depth D1–D6 Intellectual autonomy A1–A6 CALL D · waits for C Narrative & advice turns the axis result into plain prose Assemble facts + judgments ValidateRefs every citation re-checked; unresolvable ones dropped Your Mind Imprint reviewable, and a teacher can overturn it
counted facts citable evidence the model's judgments assembly & checking

Every judgment opens

What the model may cite is bounded by an index: a conversation, an event, a card, a source. What a teacher opens is the thing that actually happened at the time.

The two axes stay apart

Cognitive depth and intellectual autonomy are measured and written separately, never merged into one score. For the same concession paragraph, one axis asks how well it was done and the other asks who started it.

Thin evidence gets said out loud

If a dimension has no evidence this cycle the report says so, rather than reaching for a safe sentence to fill the gap. Nothing seen means evidence is still accumulating.

Where the difficulty is

There's a well-worn path to making an AI smarter. There isn't one for making it hold back.

None of the four below gets solved by a stronger model. Each has to be thought through as teaching first and then built — which is where we think the real barrier to entry sits.

One more thing schools ask about that isn't an algorithm: a student's conversation is handled server-side only, the browser never talks to a model directly, and no key ever reaches the front end.

01

Holding back is harder than being smart

A model's default is to complete the answer. To make it stop at the right moment, handing the thinking back has to become something it can actually invoke — a card a student will fill in and be better for. Without the card, restraint is just a line of prompt nobody obeys.

02

Support has to change by stage — and then withdraw

The same move flips from help to ghost-writing when the stage changes, so each stage's boundary is defined separately. The harder half is withdrawal: scaffolding is meant to come down, so the metric we watch is a card being summoned less and less while the question it packaged starts appearing in the student's own prompts.

03

A comment has to be checkable

“His argument has improved” is something no parent or teacher can check. So every judgment carries a real fragment: click and there's the conversation, the card, the source. A judgment that can't point at evidence goes unsaid — which is why a report will sometimes read “not enough evidence this cycle”.

04

The five feed each other

When a card should be offered is decided from process records left by real projects; how the rubric reads a level comes from seven international mark schemes and three decades of empirical work; and what the report reads is the trail those same cards leave. Any one piece is copyable. Having all of them hold at once — and improve together — is the hard part.

Start a real inquiry

No diagram beats walking it yourself.

The demo runs exactly the five above: how a course waits for her, when a card gets handed over, when a sub-agent takes the load, and how the report ends up tethered to evidence.