As long as biological entities reproduce imperfectly in environments with selection pressures, they DO have to 'stick' to evolution.
Yes, that's the abstract process:
Design Encoding -> Realisation
Realisation -> Environmental Test
Test Survival -> Propagation
Progagation -> Design Encoding Modification
I'm just flummoxed at your stubborn refusal to acknowledge the difference between the process of a human designing and modifying an object and evolution.
Then you have not undestood what I have been saying: YOU are saying the differences represent a fundamental PROCESS difference, I am saying they represent a process CONFIGURATION difference.
Now the different thing you might identify in the human design process is the use of a priori models to reason about the effect of a design change
before that change is realised. It is, however, still
sub-ordinate to the consequences of that design in the wider world and that's because we use models that
abstract away many of the details in order to reason about a change of fewer variables. Nature doesn't abstract - it doesn't care about abstractions - as such if some variable you didn't consider was actually consequential then your design will not behave as expected. This happens all the time in engineering.
When it does either we accept the fundamental limits of our abstraction or we try to formulate a new abstraction to better deal with that particular variable configuration and change so we can reason about the consequences in a computationally efficient manner.
This is why I said earlier that an evolutionary algorithm represents a highly general problem solver: its power come from its ability to 'consider' all variables in whatever 'world' it is placed in. It's weakness is the time penalty and the lack of guaruntees of optimal solutions. Deterministic problem solvers can produce certain guaruntees and if they're P class they're fast. Which is good - but it's not always possible and generally the difficulty of finding such algorithms increases with the complexity of the problem domain.
Now it's because of the way our intelligence works that we seek to solve problems in the later way and, futhermore, we tend to evaluate solutions by working backwards to try and synthesise simlpe and modular explanations. So, as has previously been explified, a human engineer probably wouldn't design the mammalian eye as it is - but he might do it like the squid eye. The problem if you want to assert that human design processes achieve things evolution cannot is that you will occasionally see things that, to a human, appear to be evidence of 'intelligent' design because they are congruent with how a human would design them.
The way you guys are going about it you cannot explain these things evolutionarilly so you are leaving the door wide open for ID guys to find examples of such things and then bark out, "SEE!? This must have been designed by an intelligence!"