I'd like to see this given a bit more structure, honestly. What occurs to me is constraining the grammar for LLM inference to ensure valid C89 (or close-to, as much can be checked without compilation), then perhaps experimentally switching to a permuter once/if a certain threshold is reached for accuracy of the decompiled function.

Eventually some or many of these attempts would, of course, fail, and require programmer intervention, but I suspect we might be surprised how far it could go.

I don't expect constraining the grammar to do all that much for modern LLMs - they're pretty good at constraining themselves. Having it absorb the 1% of failures that's caused by grammar issues is not worth the engineering effort.

The modern approach is: feed the errors back to the LLM and have it fix them.

I suspect the end goal is to have suspend/resume (or some analogue) be userspace standard library functions used to implement Io.Evented. It'll end up being a userspace implementation like you see with many of the C coroutine libraries floating around on GitHub (minicoro, llco, etc).

Edit: Looking at the working prototype of Io.Evented, this may not be true actually. Perhaps this is the domain of a 3rd-party library except with stackless coroutines?

Another thing you may want to pay attention to, however, are the current proposals for getting evented IO working on WASM -- namely, stackless coroutines as a language feature: https://github.com/ziglang/zig/issues/23446