lingo.lol is one of the many independent Mastodon servers you can use to participate in the fediverse.
A place for linguists, philologists, and other lovers of languages.

Server stats:

61
active users

#softwareengineering

3 posts3 participants0 posts today

After 6 months of heavy-use and evaluation of AI coding assistants, I've gone back to a non-AI neovim and VSCode setup. Even after a week I feel more productive. I've learned a few topics deeply by reading docs, didn't feel like I was wasting time going in circles and it honestly feels like a weight off my shoulders.

I've also learned where the checks and balances need to be in our CI system to mitigate work that may be written in part by AI tools.

I can't see any future in using AI to write code. Due to the nature of LLMs, it's never going to be able to produce code that doesn't require as much time to review as it does to write by hand.

I still occasionally use ChatGPT to ask questions but I've noticed my usage of DuckDuckGo has returned to almost normal levels too as a side effect.

I have to do these things as part of my job is to guide and advise our engineering teams. I need to be in a position to talk about these things from a perspective of experience. But it's been a tough 6 months. Glad it's over.

#ai#llm#software

#softwareEngineering #computerScience #programming #commonLisp #history #essay #reading of Whither Original Thought #transcript from @kentpitman
screwlisp.small-web.org/show/k
with a link to its beginning in the interview.

I chose to hand-transcribe it to remove (and probably introduce) discontinuities.

A central theme is the tyrannies of programming libraries and the question of having exhaustive knowledge of your increasingly numerous dead forebearers.

Read it yourself (and/or listen finally).

screwlisp.small-web.orgLiteral transcript Kent M Pitman’s Whither Original thought previously unpublished essay

Don't force software developers into using and producing #AI tools. If you want prompt engineers, get prompt engineers, not software developers—these are two vastly(!) different professions (for a reason!).

Otherwise, the industry will beg for coders in a decade, if even that far into the future—mark my words. Because they will either have quitted, unlearned or ever learned coding.

New to programming/software development?

Can I help?

I can walk us through using git, a popular source control tool. I can demo how use an IDE (integrated development environment) for debugging and writing code.

Live, now and for next 4 hours from this post. DM me.

Over 20 years experience here and wanting to give back. Free, just wanting to lessen the ick in the world.

Replied in thread

@profoundlynerdy just because something has types doesn’t make it Haskell-like. Haskell’s type system is in the family of Lambda Calculii (the “Lambda Cube”) which is called “System-F”.

I don’t know much about Raku, but it seems to me to me to be a bit more similar to TypeScript. And what differentiates TypeScript from other languages: it takes a horrible programming language like JavaScript and makes it less bad by giving it a type system, likewise Raku makes Perl less bad in the exact same way. (Sorry, I’m not trying to be impolite, but JavaScript and Perl are objectively, truly awful, horrible programming languages.)

So I see both Raku and TypeScript only being useful to a company buried in the technical debt of a hugely profitable production application that was very unwisely written in a dynamically typed language (Perl or JavaScript), which then unfortunately grew to millions of lines of code, and now it can’t be maintained by anyone, and it could never possibly be rewritten from the ground-up in a good programming language like Haskell for any reasonable sum of money. So Raku and TypeScript both offer a half-measure solution to that problem: make the maintenance of horrible computer code a bit easier with a type system.

Haskell was never intended as a fix for horrible code, it took a really good experimental programming language called Miranda and turned it into something that you can use to do real, practical software engineering, and it does it better than any other language ever invented. You write a system in Haskell because you know up front that you want it to be stable and maintained in a cost-effective manner for decades.

Zig is not similar to Raku or Haskell. It is more analogous to what Scala does for Java. Java is already statically typed, but Scala’s type system is better, and it’s runtime is fully compatible with Java. Likewise, Zig is fully compatible with the C language runtime, but provides a slightly different, slightly better static type checking system than the C type system. Zig also solves a bunch of other problems that C has by providing it with modern features like namespaces and modules, which makes it much easier to use than C. Zig is the perfect way to replace old C code with something more modern, but only if you don’t need it to be as rigorously correct as Rust. I think Zig would be a nice language to use to replace non-safety-critical front-end libraries like Gtk, or maybe for things like game engines.

I just discovered I got blocked by somebody whose opinion and posts I highly value.

Since can't imagine (or remember) which interaction might have triggered this course of action, I tried to reconstruct our last interaction(s) from web searches, and discovered the following:

(a) This is practically impossible to find past interaction with specific accounts.

(b) A couple of other accounts *I* have been following formerly which now block me without me remembering any untowards interaction with them.

This all makes me unexpectedly sad.

I think I'll put in another Mastodon moratorium for an indeterminate length of time.

Looking back, Mastodon did not really work for me, with what I'd term "meaningful" connections. This is very likely the way my mind works.

If anybody can recommend communities related to #lisp, #agile, #softwareengineering or #infosec, I'd appreciate if you could reply or DM. I am still missing connection there (after 2.5 years of Fedi, which should tell me something).

Said this to a colleague in work Slack just now:

"I do have to say (if we can pause to pat ourselves on the back here) that I think you and I (and [x]) have done a really good job at designing things to be simple, only adding complexity when really necessary, and also fighting like hell to *avoid* adding complexity."