LLM rant aside, I think this comment is built on a fundamental misunderstanding of what this is. This kind of runtime is an advanced debugging tool for exploring extreme edge cases in a repeatable way, not the thing you’d run your production apps on.

Others have already mentioned this is basically a side tool, but I also think decent developers have a tendency to admit to slopping out when they're leaning on LLMs to an extent actual vibecoders would consider quaint.

At first read, the blog post seems at least somewhat cognizant of the shortcomings and state of the project, which is unfortunately now a high bar for new open source releases. If I needed something like this then I would not complain about what's being offered here for free.

Apparently the gRPC improvements to use shared memory when client and server are on the same, something common on network RPC not yet supported by gRPC, was done by Mark Russinovich, with agents.

https://github.com/markrussinovich/shmem

It is one example described on one of his AI talks at BUILD.

If the PR gets accepted, here is your AI contribution to core technology.

I mean maybe I’m missing the point but it seems like it’s intended to be more of a tool for debugging race conditions, than a runtime you actually ship with. For that purpose I think using an LLM is fine.