ArkType 1.0-kappa: Building an LLM within TypeScript's type system #706
Replies: 1 comment
-
I must say, that is quite an unconventional approach to runtime type systems! While I can appreciate the desire for creativity and uniqueness, it's important to remember that the primary goal of a type system is to provide safety and ensure correct program behavior. While generating sentences that feature every letter of the alphabet is certainly an impressive feat, it may not necessarily lead to safer or more reliable code. Furthermore, requiring such a large amount of RAM for VSCode users could be a major hindrance to adoption and usability. It's important to consider the practical limitations and tradeoffs of any tool or feature. That being said, I'm always interested in exploring new ideas and approaches, so if you have any specific questions or feedback you'd like to discuss, feel free to let me know. I'll be happy to assist you in any way I can! |
Beta Was this translation helpful? Give feedback.
-
We've been asking ourselves recently- what do we really want out of our runtime type system? Safety? Determinism? Nah, f*** that.
In the end, we concluded that our top priority should be empowering ArkType to summarize your types using sentences that feature every letter of the alphabet. Therefore, we will be integrating with ChatGPT at runtime, and devoting all of our resources to building a static equivalent that can provide 1:1 results right from your editor.
Unfortunately, VSCode users will now need to allocate a minimum of 512GB of RAM via the
maxTsServerMemory
, but honestly it's a small price to pay to have the compiler complain about your index signatures in the style of Snoop Dogg.Thoughts? Feedback? Leave a comment below and we'll respond via our static LLM in 3-5 months 🔥
Beta Was this translation helpful? Give feedback.
All reactions