Human-led AI collaboration
Brainstorm with AI. Keep the thinking yours.
Use AI as a thinking partner without giving up authorship or initiative. Start with your own idea, make the key choices, test what happens, and keep a visible trail of how your thinking changes.
Beyond the chat
A conversation is a line. An idea has depth.
Each message is a flat view of a larger idea. As you and AI question, test, and revise it, important insights land in a structured map—preserving what began with you and showing how the idea grew.
A worked example
Try the Newton inquiry
Follow Newton from a falling apple to a question about the Moon. The demo shows how conversation, evidence, and revision become a traceable line of thought.
The partner's side of this conversation is written in advance. No AI model runs here, and nothing you do in the example leaves your browser.
Why it's built this way
The theory behind the design
This is not a proven method or a promise — it's the working theory the design follows. A few plain commitments on the surface, and three deeper ideas underneath.
- Ideas keep their provenance: a kid's own thinking stays the source of truth, and AI can illuminate it without owning it.
- Revision counts as learning: changes, disagreements, and abandoned paths stay part of the story instead of being polished away.
Deeper down, three ideas carry the design:
-
Levels of abstraction — leading more with less
Strong thinking climbs a ladder of abstraction. As AI gets better at filling in details, the human contribution moves up the ladder: name the goal, the strategy, the assumption — and steer a lot of machinery with a few clear ideas. The map exists to make those few ideas explicit.
-
In the age of AI, abduction matters most
Once the assumptions are given, deduction — and much of induction — can largely be handed to machines. What stays rare is abduction: the leap to a new explanation, the “what if it works like this?” Einstein's relativity began as that kind of leap, not as a longer calculation. So the design keeps the guessing, and the naming of assumptions, with the kid.
-
Active construction, not a mental hand-over
Creativity and authorship live in the act of keeping track — knowing why each piece is there. When thinking is handed to AI, even just mentally, that tracking quietly disappears. So nothing lands on the map by itself: reading never writes, and every idea is placed by the kid who owns it.
An invitation
Think it through with us
If this direction speaks to you — as a parent, a teacher, or a researcher — we would love to compare notes. Try the example above, then tell us what you noticed.