off-topic

mloughlin 2026-01-13T10:39:41.003679Z

Are there any recent (post "Maybe Not") essays or talks about the limitations of type systems? Maybe I'm old and cranky but the "static types are always better" crowd seem to be teaching the next generation like it's a fact rather than opinion

2026-01-13T12:20:09.013059Z

FWIW GPT can makehttps://chatgpt.com/share/69663841-a038-800a-a3ee-8279fe126e75

👎 2
👎🏻 1
respatialized 2026-01-13T14:28:42.698649Z

https://www.hillelwayne.com/post/this-is-how-science-happens/ Hillel Wayne's newsletter and essays come back to this topic from time to time. Here's a writeup of a paper that attempts to empirically measure the impact of language on whether a commit introduces a defect in large-scale codebases. The results table is somewhat instructive. According to their data and defect model, Typescript seems to drive the largest reduction from the baseline defect rate. However, Clojure performs quite well! It is the runner-up and outperforms languages with powerful type systems like Scala and Haskell, and strongly outperforms other popular statically-typed languages (golang, C#, Java). Ruby also appears to outperform these languages.

respatialized 2026-01-13T14:30:33.664939Z

I would also say that "proportion of commits that introduce a defect" doesn't capture every variable relevant to software engineering as well. A static type system can prevent you from committing broken code, but merely measuring the code that gets committed may leave out other variables, such as the time it takes to get a build working again after you broke something in the type system, which is likely happening prior to the actual commit.

2026-01-13T15:28:16.091139Z

My idea when sharing the GPT summary was not to share any truth oc. It was just to summarize the arguments that have been repeated again and again in different talks, essays, etc about the cons of static typing as perceived by programmers which would resonate or not with different people experiences (they do with mine). IMHO this is one of those cases were trying to do science misleads more than it helps, because the science will focus on a couple of metrics (like defects rates) which while an important aspect is far from the whole picture (there is fun, motivation, different contexts and domains, etc), and tends to push people away from accepting there are trade-offs that needs to be taken into account in each project, team, etc under the "this has been proved by science to be better".

2
2026-01-13T14:49:58.457949Z

I got into lisp working on a open source ai project that im forking till febuary i learned lisp using the wizard book and lisp and hylang website

1
2026-01-13T14:50:43.960739Z

So old techniques are still solid