The vibes:
be the reason people are afraid to litter in the forest.

Review my code, please.

Did you know that software written by human is proprietary by default?

Unless it re-licensed under other license by the author.

Making Scheme code uglier until it is impossible to read it.

#scheme

During last a couple of weeks I've spent enormous amount of time reviewing Scheme code. My top reasons of miserable and painful review experience:

- Fucked up POSIX API.
- Text representation of the code.
- Lack of types and effects declarations.
- Verboseness of Scheme.

However, it's almost an actionable observation.

I have more exact issues/use cases written down and I have thoughts on how they can be fixed.

The only problem is: it's quite a challenge to implement those "fixes" :D

I merged a PR, and now the commits are not found :blobcatscared:

Okay, friends, we don't need smarter and faster LLMs. We don't need better harnesses. We don't need to produce more code.

Review is the bottleneck.

We need to improve and simplify the languages, system APIs, review and verification processes.

We always needed better foundations, but today those issues shines like diamonds in the sun.

And I bet that "AI" won't magically solve it for us.

Woah, the Unix is so cursed. I'm suprised how the applications and kernels still work somehow with all that dup2, fd's, fork-child-exec-and-whatnot crap under the hood.

It's so overcomplicated, It's so brittle. It's a torture, not a programming.

Yeah, you can guess that I'm reviewing some Scheme code with these tasty Unix APIs calls.

Conciseness is important. Too little and it's verbose unbearable noisy mess, too much and you won't decrypt what is going on.

The first thing I feel, when I see 50 commits PR I need to review:

Rocq-solid Software - a formally verified/proven piece of a program.

Example usage: I want both the kernel and userspace programs to be rocq-solid.

P.S. Doesn't matter what proof assistant you used: #coq/#rocq, #IsabelleHOL, #agda, #tlaplus or whatever.

I have one more reason to make code content-addressable.

Right now people have became much more capable in producing code with the tools like LLMs, but it definitely increases fragmentation and isolation in the communities.

However, we still have a lot to share and talk about and if the code was content-addressable, we could literally see that other people work on the same ideas as we do (even across the boundaries of one language) and communicate and cooperate on these matters.