Main idea

What a competitor publishes — a module, a screen, an integration, an AI agent — is the output of a line of reasoning that does not itself come out: which customer it targeted, which outcome it sought to produce, which alternative it was fighting, which internal constraint led it to that shape rather than another.

A feature inventory therefore collects the part of the work that carries no information. Two products can address the same need through very different answers, and two identical answers can come from opposite lines of reasoning. Reproducing the answer without having reconstructed the reasoning carries over nothing of what made it right: it is reproducing a gesture.

Why it matters

This explains why copying fails without anyone being able to blame poor execution. The copied feature works, it ships, and it does not produce the effect observed at the other company — because the effect came from the target customer and the alternative being fought, not from the shape.

It also gives the right question to ask when facing a competitor's release: not "what do we need to catch up on?" but "what problem is this trying to solve, what gain is promised, what friction is removed?" — a question you can answer without copying.

Nuances and limits

The reasoning is not always reconstructible, and it is not always deep: a competitor's feature can be the result of a big customer's request, a technical opportunity, or a scheduling trade-off. Systematically attributing strategic intent to what you observe is the symmetrical error to copying.

And some answers have become market expectations: reproducing them is then not a strategy but a condition of entry, which is a different and legitimate decision.

Open questions

• How do you tell that a line of reasoning reconstructed from the outside is plausible, rather than projected by the person observing it?