They could argue that by viewing the copyrighted code/implementation, you could effectively infringe by (even subconsciously) writing the same/similar code.

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.

Ah, so it's not the act of disassembling that's the problem, but that you're infringing on the original code's copyright? That makes sense, thank you.

You may have come across this concept already, but this is where clean rooms come in.

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.

Yep, I knew of clean-room reimplementations, I was just wondering whether decompiling is somehow in itself illegal.

But disassembling/decompiling doesn't give you anything like the original code!

It gives you a derivative work of the original code

That kind of depends on the language, but it's a fair point. I think it might only matter that the general algorithm/solution is the same, not the lines of text themselves.