None of this addresses my point in the context of your original claim about a radio always being a radio.
The point is about limitations of variability in hierarchical information systems, the radio being merely a simple proof of principle.
Where a range of variation is inherently limited by the system that supports it
Maybe a better example, since it involves digital information:
You can throw random values at the text parameters here,
some will work better than others, but you have a reasonable shot at a viable outcome, because the system specifically limits the options to viable ones
Within this software architecture, 'random mutations' work fine, and can even be used to search for optimal configs- as your experiments also demonstrate. But of course you understand why you can never write a new program this way- you can't simply extrapolate these minor variations into a 'macro' change into a new software application
Now try throwing random values at the source code in your operating system-
the chances of improving your operating system are not zero, they are just vanishingly small- because the options are not constrained to viable ones, so there are an infinitely greater number of ways to crash the system rather than improve on it.
This is exactly the same with DNA- it's a hierarchical digital information system which is uncannily computer-like. sure you can throw random values at eye color control genes, they are set up that way- but you simply cannot extrapolate these superficial variations into macro evolutionary ones.
in short; biological adaptation is a design
feature, not a design
mechanism