There's merits to this claim if you're indeed implementing some advanced, niche algorithm but it definitely wouldn't apply here as all he's doing is calling HTTP APIs, a very generic and common thing to do.
One person views the "contaminated" decompiled code and writes a specification. A separate person writes the code based solely on the specification. This is an accepted method of demonstrating that there is no infringement.