Program synthesis is cool. Learn more about it!

Deterministic methods aren't as useful as they sound… a complete
specification that can generate a program **is** the program, and is
usually just harder than writing the program that follows that
specification - though it does act as a second implementation that can
precisely validate the first.

If we recognize that we're okay with some undefined behavior, we can
describe and synthesize programs with a series of property tests; this
can be useful for checking work or if we have a human in the loop to
validate everything we're generating (but in this case, we want
nondeterminism, deep learning and these sorts of things).

[Question Selection for Interactive Program Synthesis (PLDI 2020 - PLDI
Resear…](https://pldi20.sigplan.org/details/pldi-2020-papers/65/Question-Selection-for-Interactive-Program-Synthesis)
[Writing a Test Case Generator for a Programming
Language](https://fitzgeraldnick.com/2020/08/24/writing-a-test-case-generator.html):
Generating test cases (fuzzing?) for WASM.

[Why care about Program
Synthesis](https://www.zinkov.com/posts/2019-02-17-why-program-synthesis/)