REVERSE
Do not copy pixels. Extract transferable principles.
Decomposes a reference product into observed behavior, flows, information architecture and system hypotheses, separating transferable principles from patterns to avoid.
What remains when the system finishes?
The output is not a pile of slides; it is a usable work package that directly feeds the next decision or implementation step.
E0–E4 evidence-graded observation log
Product model, user flows and information architecture
Technical system hypotheses — explicitly separated from observation
Adopt / conditional / avoid principle set
What is in, and what is not?
One primary reference product and focused critical flows
Public surfaces; authenticated surfaces only with client-provided access
No source-code extraction, access-control bypass or visual copying
Produces product principles; FORGE defines our product architecture
Why does Linear feel fast?
We decomposed Linear's publicly documented creation, Peek, keyboard, search and inline-edit behavior into principles without copying screens.
OUTPUT / Observation → evidence → principle → transferability → ADOPT / CONDITIONAL / AVOID.
When should you use this system?
You want to understand why a product works well
You need behavioral decomposition instead of a feature comparison
You want strong reference evidence before FORGE
Close this bottleneck, then move only to the stage you actually need.
Possible next system: FORGE / Product architecture. Skip it if that bottleneck is already solved.