Every answer checked
The answer key is not a vibe.
Every exercise in Felt goes through a validator before you ever see it. Maths is evaluated symbolically, code is run, structured answers are matched against rules. When the validator can't be sure, the exercise is flagged, not shown to you as fact.
What most AI study tools do
- A model writes the question and the answer in one go.
- Nobody checks the key. If the model is wrong, you learn the wrong thing.
- Wrong answers get a generic explanation, or none.
What Felt does
- An AI author writes lessons and exercises.
- A separate validator checks every answer against rules, not the model's mood.
- Exercises that fail the check never reach you. Wrong answers get an explanation for that exact task.
| Subject | How the answer is checked |
|---|---|
| Maths, physics, chemistry | Numeric and symbolic evaluation, units and tolerances, step-by-step problems scored by method |
| Programming | Your code is run against test cases |
| Economics, graphs | Diagrams you move are read back as data: shifts, intercepts, equilibria |
| Definitions, structure, labels | Matched against rules and accepted variants, not a single string |
| Anything the validator can't verify | Flagged as unverified, never shown as fact |