For students
Guidance without answer dependency
Hints respond to the reasoning step that is missing, then fade as independence grows.
AI for deeper learning
Physics is not a list of answers. It is a way to think.
Lattice turns every problem into a conversation about evidence, assumptions, and models—helping students build judgment alongside subject knowledge.
- First curriculum
- Cambridge IGCSE Physics
- Core outcome
- Transferable reasoning
- Current stage
- Concept prototype
A trolley accelerates even though the pulling force stays constant. What changed?
Your claim
The trolley became easier to accelerate.
Test the idea
Which quantity in F = ma could have changed? Choose one and explain what evidence would support it.
Identify the observation
Propose a mechanism
Test against evidence
Revise the model
Most learning systems optimize for the next answer.
Lattice is designed around the next better question.
The learning loop
One problem. Four acts of reasoning.
A session adapts to how a student explains, tests, and revises—not only whether the final answer is correct.
- 01
Observe
Separate what is known from what is assumed.
- 02
Model
Turn an intuition into a claim that can be tested.
- 03
Challenge
Use counterexamples, evidence, and uncertainty.
- 04
Transfer
Apply the reasoning pattern in a new context.
Designed for the classroom
Student agency. Teacher visibility.
For teachers
Misconceptions made legible
Class views group recurring reasoning patterns so the next lesson can address what actually blocked learning.
For schools
A measurable reasoning curriculum
Progress is tracked through transfer tasks, explanation quality, and productive use of feedback.
What we intend to prove
Better thinking should leave evidence.
These are pilot hypotheses—not achieved results. The first studies would test whether the product improves independent reasoning and helps teachers intervene earlier.
Can a student reuse a reasoning strategy in an unfamiliar problem?
Does reliance on hints decline while explanation quality improves?
Can misconception patterns be recognized before the next assessment?
Concept prototype