Observe → infer → execute
What happens next
is learned.
Turn observed API behavior into an executable model. Change inputs and explore inferred state rules, including memory a public snapshot can miss.
A public snapshot can miss hidden memory.
Native comparisons are separate recorded evidence, never live measurements.
Loading the portable session.
01 / Experiment
Write a history
Add an action
Choose an action to see its declared argument types.
A refused action keeps its previous state. Unsupported inputs are explicit; prediction does not establish native equivalence.
02 / Follow the behavior
Every transition, exposed
Selected step
Start with a history
Inspect the inferred rule
Keep the evidence separate
Learned does not mean unconstrained.
The interface and argument types are declared by a person. Elementary arithmetic, sequence operations and UTF codecs are supplied grammar. The model's transformations, guards and distinguishable modes are inferred within that grammar.
Recorded native comparisons No receipt
No native oracle runs in this studio. Imported evidence is unsigned data.
Build your own model with the Python CLI.eidryl --help · learn → check → export