This language is not interesting for what it is, but how it was made. Apparently the guy put this Claude into a loop for 3 months [1] and this popped out. This made him one of the most prolific users of Claude, using tokens at a ate of up almost half a million dollars in a matter of days [2]. This caused considerable churn on the repo:
1,196 commits 12,001,326 ++ 8,639,553 --
~1000 commits over a 3 month period netting 3.4 million lines. The git history alone is over 400MB after just that short period of time. Go's is smaller. What's the codebase look like?
--------------------------------------------------------------------------------
Language Files Lines Blank Comment Code
--------------------------------------------------------------------------------
JavaScript 1444 1243716 25330 750735 467651
Markdown 981 294632 49397 0 245235
Zig 380 226254 36039 21445 168770
TypeScript 807 172719 7893 95437 69389
JSON 196 47506 6 0 47500
HTML 448 38447 227 8 38212
Plain Text 44 29594 1481 0 28113
YAML 5 2344 228 167 1949
Python 5 2151 375 130 1646
Toml 30 1917 315 328 1274
C 9 107169 156 106074 939
C/C++ Header 6 961 134 96 731
CSS 3 809 121 19 669
VimL 4 775 102 98 575
Lua 2 471 60 34 377
XML 1 359 69 28 262
Lisp 5 303 36 36 231
Typescript JSX 1 497 63 253 181
Ruby 1 244 49 32 163
PowerShell 2 207 28 19 160
Makefile 1 213 35 24 154
PHP 1 156 13 19 124
Docker 1 211 42 54 115
Bourne Shell 6 136 33 15 88
Kotlin 1 78 13 0 65
Protobuf 1 30 7 2 21
Batch 3 15 0 9 6
--------------------------------------------------------------------------------
Total 4388 2171914 122252 975062 1074600
--------------------------------------------------------------------------------
Wat? Like the another comment said, the language is essentially Go with the keywords changed. You could get a better result just by forking the Go repo and changing the keywords, rather than spending millions of credits and months on LLM churn, which results in a completely unmaintainable codebase of unknown quality.
Anyway, the results are bad. The language is what it is, I can't judge it because I don't think it's trying to be anything more than a joke. But the process is just... awful, and I don't think letting the LLM just churn is a model for sustainable language development.
[1] https://ghuntley.com/cursed
[2] https://x.com/GeoffreyHuntley/status/1950243941917307338
[3] https://github.com/ghuntley/cursed/graphs/contributors
9
6