I know, which seems absurd to me. Like, TUIs are 1) a solved problem, b) performant, and c) right there, on every single computer in the world. (Someone please correct me, to point out that there's some weird OS somewhere that doesn't have a shell, so I can say "Yeah, but does it run Claude?")
Is there an engineering-based reason, or is it that the AI knows React, so it's easier to vibe, and damn the user experience?
I'd say websearch and webfetch tools are very important but missing here. Great direction though, I also wonder about how much bloat could be eliminated...
Claw-code (reached 100k stars on github and I have made an article about them[0]) are currently re-writing claude-code in rust (they have already ported it to python)
This is built by studying the leaked source.
"Clean-room rewrite" doesn't mean much when
you've read the original first — that's not
clean-room, that's copying with extra steps.
This is theft, plain and simple.
The stars on the original claw-code repo were
earned through the leak, not the work. Riding
that wave with another clone doesn't make it
original engineering.
Also worth noting: this wasn't submitted as
Show HN — it's being promoted through comment
threads instead. Make of that what you will.
Real open-source respects both the letter
and the spirit of IP law.
I added a new track flight function and you can track any airline in flight.
How much of this was written using / by Claude? Not hating or snarking, BTW, just curious.
Why (especially if the answer above is 'yes' / 'most') do we collectively think Anthropic built such a heavy interface?
all of it, and I think the answer is 1. feature creep 2. a lot of nice-to-haves. the TUI alone is responsible for a lot of bloat
Even that TUI is written in React.
I know, which seems absurd to me. Like, TUIs are 1) a solved problem, b) performant, and c) right there, on every single computer in the world. (Someone please correct me, to point out that there's some weird OS somewhere that doesn't have a shell, so I can say "Yeah, but does it run Claude?")
Is there an engineering-based reason, or is it that the AI knows React, so it's easier to vibe, and damn the user experience?
comes out to about ~1,000 lines in python as well https://gist.github.com/seanwevans/c9d44aadae9690ec2325bc940...
Claude code written for basic on your Apple II, because it'd be so retro.
I'd love to see it
I'd say websearch and webfetch tools are very important but missing here. Great direction though, I also wonder about how much bloat could be eliminated...
agree, I was more focused on bloat than pure utility
ok this is actually impressive, really. idk why no one commented yet
That sure is a lot of lines to avoid using getopts...
There seems to be only one possible option, --resume, and only one line is dedicated to handling it...
I am now waiting for, Claude Code rewritten in assembly project :)
(Maybe it can be a good project for anyone who works low-level)
we’re going to have a bunch of CC rewritten in <obscure language>
Waiting for Rust.
Claw-code (reached 100k stars on github and I have made an article about them[0]) are currently re-writing claude-code in rust (they have already ported it to python)
Its worth following: https://github.com/instructkr/claw-code
[0]: my article: https://smileplease.mataroa.blog/blog/claw-code-reaches-100k...
This is built by studying the leaked source. "Clean-room rewrite" doesn't mean much when you've read the original first — that's not clean-room, that's copying with extra steps. This is theft, plain and simple.
The stars on the original claw-code repo were earned through the leak, not the work. Riding that wave with another clone doesn't make it original engineering.
Also worth noting: this wasn't submitted as Show HN — it's being promoted through comment threads instead. Make of that what you will.
Real open-source respects both the letter and the spirit of IP law.