Frank Laub

A new idea could make this possible, will revist soon.

Next step is to submit a PR with changes to `risc0_zkvm::serde` then let our CI chew thru it with lots of examples to see if we run into any issues....

Can you provide more context as to how these are used and also how they might be implemented?

I think we need to have some sort of process that decouples what's happening on the HEAD of main vs what the latest release is. We might temporarily have performance...

I'm going to get the datasheet to be updated, but I still think we should have a separate release process for the datasheet for public consumption.

I think we should recommend this instead: https://github.com/risc0/toolchain/releases/tag/2022.03.25

Also, there was a previous attempt to make this easier here: https://github.com/risc0/risc0/pull/205

This might just be that the PATH needs to include `/usr/local/cuda/bin`.

Under `no_std`, there doesn't seem to be a way to avoid some sort of mechanism to mark what the entrypoint is.

I'm thinking we should long term try to move towards using `std` for most things, assuming `std` doesn't cost us much overhead.