Commit before explanation
The system needs to see the student's first move—not the polished story they can tell after seeing the answer.
Evidence-led learning, built first for LSAT
Understanding is not a feeling.
It is something you can show.
Natalie is building a learning system that makes reasoning visible—before, during, and after an answer is chosen.
01 / THE LEARNING LOOP
V1
not
“I get it.”
but
show me.
CommitChallengeTransfer
01The thesis
Recognition can look like learning.
It is not the same thing.
Once students see an answer, they can often retell its logic. They feel that they have learned it. Change the surface details, though, and the same reasoning failure returns.
Natalie begins at that gap. It does not accept a fluent explanation or a correct selection as proof. It asks for a commitment, locates the exact inference, and tests it in a changed situation.
02A different standard
A good answer is only the beginning.
Make the inference precise
It follows the specific step that holds or fails: a condition, a distinction, a causal bridge, or an eliminated alternative.
Require transfer
The student works through a fresh example that cannot be solved by repeating the earlier wording.
03The learning loop
Designed to surface
the next real question.
At each turn, Natalie gives students a small, purposeful set of moves and one way to say something the system did not anticipate.
-
01
Present
A short task with one clear learning objective.
-
02
Commit
Choose or state an answer before an explanation appears.
-
03
Locate + probe
Identify the reasoning step, then pursue its weakest point.
-
04
Challenge
Test a tempting alternative, contrast, or counterexample.
-
05
Transfer
Use the reasoning in a meaningfully changed example.
-
06
Record + recommend
Keep the evidence and choose what deserves to be tested next.
A new interaction paradigm
Not a replica of a person.
Not an open-ended chat window.
Natalie is candidly a machine. Its value is not a simulated relationship; it is a disciplined interface for learning. Text-led dialogue gives a student room to think, while structured choices make productive questions visible before the student knows how to form them alone.
Normal turn
Up to 3 purposeful moves
+ a custom response
04V1 focus
Start where precision
can be observed.
FIRST LEARNING SLICE
CONDITIONAL REASONING
“Only if”
The first unit asks students to translate a conditional rule, distinguish necessary from sufficient conditions, and reject the invalid reversal in a new context.
It revealsPrecision, inference, rationalization, transfer
It requiresObservable evidence—not self-report
05Built to stay rigorous
Adapt the presentation.
Keep the standard.
The system can adapt pacing, scaffolding, example choice, and the next question. But each adaptation must point back to observed evidence. A student profile is a revisable hypothesis—not a verdict—and later work is meant to challenge it as much as confirm it.
Natalie