How do we maintain best practices when the compiler outputs a different result for the spec at any given time? How do we obtain reproducible builds? Do we pin to a specific version of our compiler (ie, snapshot of the model; is this possible anywhere except local currently?), and vigorously test changes after any updates in our "toolchain"? How do we have control over our "toolchain" (again, apart from local), especially when said "toolchain" can, for all its users simultaneously, fold to political pressure from state regimes? And, if the code generated by LLMs is the build artifact, why is it now okay to check the build artifact into source control?

There may come a day when we, as an industry, decide that simply doing it by hand is more expedient when it comes to resolving urgent production issues. We may not know the pain we are causing ourselves until well into the future when it has become too much to bear without a visit to the proverbial doctor.