> Its runtime proofs are based off an abstraction of the algorithm suitable for direct manipulation by proofs rather than the actual implementation in code.
What is the difference? You are aware that the code is also only an abstraction, right?
The difference is, that proving something about an abstraction doesn't prove, that you made no mistakes when translating that abstraction into the actual code running, and therefore you have not proven anything of value about the actually running code.
The difference is, that proving something about an abstraction doesn't prove, that you made no mistakes when translating that abstraction into the actual code running, and therefore you have not proven anything of value about the actually running code.