Users appreciated the Haskell veteran's thoughtful proposal to switch to Lean for harnessing AI theorem provers, praising the post as well written.
Based on 2 visible X reactions from 2 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@avi_press @satnam6502 I think it came through in the text: very well written and thoughtful, thanks!
@RanjitJhala @satnam6502 Appreciated ❤️
This article by @avi_press describes why he switched some of his company's (new) code from Haskell to Python, and he makes a lot of great points. The relative value of the Haskell type system has diminished in this AI world, and the slow build times hurt more than ever. It would be interesting to imagine a world where they had switched instead to Lean, retaining many of the advantages of Haskell but now gaining significantly more value from the type system e.g. being able to use AI theorem provers to prove formal properties about your code (or aid the AI synthesis of code with formal properties). Build times may still be an issue, but I suspect sufficient mitigations could be developed. And before you kick my head in, I've been using Haskell for 36 years and I am a huge fan. https://avi.press/posts/2026-07-10-after-7-years-in-production-scarf-has-reluctantly-moved-away-from-haskell.html
This article by @avi_press describes why he switched his company from Haskell to Python, and he makes a lot of great points. The relative value of the Haskell type system has diminished in this AI world, and the slow build times hurt more than ever. It would be interesting to imagine a world where they had switched instead to Lean, retaining many of the advantages of Haskell but now gaining significantly more value from the type system e.g. being able to use AI theorem provers to prove formal properties about your code (or aid the AI synthesis of code with formal properties). Build times may still be an issue, but I suspect sufficient mitigations could be developed. And before you kick my head in, I've been using Haskell for 36 years and I am a huge fan. https://avi.press/posts/2026-07-10-after-7-years-in-production-scarf-has-reluctantly-moved-away-from-haskell.html
@avi_press @satnam6502 I think it came through in the text: very well written and thoughtful, thanks!
This article by @avi_press describes why he switched some of his company's (new) code from Haskell to Python, and he makes a lot of great points. The relative value of the Haskell type system has diminished in this AI world, and the slow build times hurt more than ever. It would be interesting to imagine a world where they had switched instead to Lean, retaining many of the advantages of Haskell but now gaining significantly more value from the type system e.g. being able to use AI theorem provers to prove formal properties about your code (or aid the AI synthesis of code with formal properties). Build times may still be an issue, but I suspect sufficient mitigations could be developed. And before you kick my head in, I've been using Haskell for 36 years and I am a huge fan. https://avi.press/posts/2026-07-10-after-7-years-in-production-scarf-has-reluctantly-moved-away-from-haskell.html
This article by @avi_press describes why he switched his company from Haskell to Python, and he makes a lot of great points. The relative value of the Haskell type system has diminished in this AI world, and the slow build times hurt more than ever. It would be interesting to imagine a world where they had switched instead to Lean, retaining many of the advantages of Haskell but now gaining significantly more value from the type system e.g. being able to use AI theorem provers to prove formal properties about your code (or aid the AI synthesis of code with formal properties). Build times may still be an issue, but I suspect sufficient mitigations could be developed. And before you kick my head in, I've been using Haskell for 36 years and I am a huge fan. https://avi.press/posts/2026-07-10-after-7-years-in-production-scarf-has-reluctantly-moved-away-from-haskell.html
Users appreciated the Haskell veteran's thoughtful proposal to switch to Lean for harnessing AI theorem provers, praising the post as well written.
Based on 2 visible X reactions from 2 accounts; directional sample.
Ask a question below.
Published answers will appear here.