Learn with the AI. Own the judgment.
Mind Imprint is an AI workspace where secondary students do real coursework. The AI searches, questions and reviews alongside them the whole way — and never writes their prose or settles the answer for them. Every judgment they make leaves a trace, and the traces grow into a report you can read their thinking in.
Rolled out class by class · every step recorded · no composite score · no ranking
Students already use AI. The question is how.
They don't lack tools. They lack occasions to actually make a judgment. Ask the AI something and thirty seconds later a fluent, complete answer is sitting there — and whether it's right, whether it can be used, whether something better-founded exists, usually goes unasked.
The answer arrives too fast
Fluent, complete, plausible. Whether anyone stops to ask “is this true” comes down to habit.
Suspicious, but can't say why
“A glass of lemon water a day prevents cancer.” He senses it's off — but not where to start checking, or what would settle it.
Handed in, and can't say which part was his
The process is scattered across dozens of chat messages. Nobody can see what he actually thought — including him.
From too much information to too many answers
Education's hard problem always tracks whichever ability is scarcest at the time. Today that's judgment: a table full of plausible-looking conclusions, and a call to make about which to believe and which to use.
The advantage went to whoever could find the material.
Finding got easy. Choosing well did not.
A pile of competing, plausible-looking conclusions — and a call to make about which to believe and use.
Judgment is a set of concrete moves. Whether a source can be used. How far a piece of evidence reaches. Where the AI's suggestion breaks. Whether your own position needs to move. What Mind Imprint keeps is the process where those moves happen.
Three things, practised on his own coursework
None of the three can be learned on paper. They only grow inside a task long enough and real enough that the AI can genuinely help — and genuinely lead someone astray.
Critical thinking
Trace a claim to its primary source, read how far a piece of evidence actually reaches, take on the counter-case, and keep the conclusion within its bounds.
Exploring on his own
Narrow a vague interest into an answerable question, decide where to dig — and know when to stop digging.
A working line with AI
Know what to hand the AI and what to keep. And when refusing its suggestion, be able to say why.
A student can use it freely, and a school can let them
The AI writes no prose and settles nothing. There are no streaks, leaderboards or badges holding anyone here. Measurement is never used for selection, ranking or discipline, and informed parental consent comes before any of it is recorded — written into permissions and contracts, not only into a privacy policy.
“Has China made the planet more sustainable?”
One student, an 800-word essay, four surfaces. Below is the route she actually took.
A course hands her a method she can take away
A course is a workflow: watch the material, get questioned, make a call, fill in a card, move on. CRAAP, SIFT and well-formed argument all get practised on real material — a method only really lands once it has been used.
An unsourced post — and she traced it to Nature herself
The post claimed “over 40% of the Earth's greening comes from China”. She read laterally — post → NASA release → the Nature Sustainability paper — and found the denominator had been swapped. The original 42% was the share of China's own greening due to forest programmes.
The argument hits a counter-example — she writes the concession
China is the world's largest emitter. That counter-evidence doesn't sink her thesis, but it does bound her conclusion. The AI writes none of her prose — it only asks: are you going to take this on, or route around it?
Even the time she skipped is on the record
The report has two halves: facts that are counted, and judgments that are grounded. Every judgment must point back to a real fragment — the ones that can't are dropped. The two axes are reported apart, never merged into a single score.
The four places a student spends time
Each one helps him do the inquiry properly, and leaves a trail that can be checked later. The trail is a by-product of doing it properly, not surveillance bolted on top.
Writing project space
He brings real coursework: framing, sources, reading, writing, review. One continuous conversation walks it with him, and he writes the prose.
See how it's designedCourses & thinking cards
34 thinking cards and 9 disciplinary lenses. Courses teach the method; a card puts it in his hands at the moment he needs it.
See how it's designedProcess evaluation
Cognitive depth and intellectual autonomy, measured and reported apart. Every judgment opens onto the record behind it.
See how it's designedTeacher console
Teachers can finally see the process: a narrated week, a live roster, openable evidence — and a one-click override.
See how it's designedTwo axes. Measured apart. Never merged into one score.
Recording isn't enough. Chat logs, action logs and version history piled together do not become an assessment. Reading a process takes a ruler with a real lineage.
Both axes come out of three decades of empirical work on critical thinking: one reads how far down the skill went, the other reads what the disposition turned into as behaviour.
See the full assessment designHow many layers down the thinking got. Depth sets the ceiling on quality.
When nobody asked: who started the thinking, and whose judgment is it.
Autonomy without depth produces confident garbage. Depth without autonomy is a good engine rented to whoever prompts it.
Four red lines hold the AI back. Six commitments hold us back.
A product that argues about judgment has to survive being judged. Where it lines up we say so line by line; where it doesn't, we say that too.
Read every commitmentThe AI never writes the student's prose
Explaining, hinting, playing devil's advocate, flagging a broken draft — all fine. The core judgment and the body text are never produced for her.
No manipulation
No streaks, leaderboards, badges or engagement pushes. A card may be triggered automatically — opening it is always her call.
The axes are never used to select or rank
Never for admissions, ranking, discipline or staff appraisal. Never sold, never turned into a commercial profile.
When the evidence is thin, we say so
No evidence seen only means evidence is still accumulating; it says nothing about the student's ability. We'd rather say nothing than pin on a label.
We're trying to grow people who can think further with AI, and still own the judgment at the end of it.
Mind Imprint white paper · Conclusion
Walk it yourself first. Then decide whether it belongs in your classroom.
The demo runs on real material: an unsourced post, traced all the way back to NASA and Nature Sustainability. You'll see when a card gets handed over — and what the final report says.