From the Abstract Heresies blog: "Githack is a Lisp object store that uses Git as its backend."

> Next up is demonstrating the AI is more productive with better programming languages.

That actually would be pretty exciting!

But, wouldn't AI be biased toward more popular languages, since those will by nature of their popularity provide more sources for training material?

Even still, the AI could demonstrate which algorithms or maybe patterns and techniques are more productive, in the context of the popular languages.

You might be surprised. I've had an AI start spitting out none functioning python code. I've rewritten it with Go. The requirements are that every change has to pass compilation, linting and tests. Same requirements as Python. Go keeps the AI on the rails better. The compilation failing provides quick and probably better feedback than a whole dictionary of tests.

For sure. Its code output is definitely not to be fully trusted. Oftentimes, you can point at where it made a mistake and it usually fixes it too.

That’s not necessarily a bad thing.

As an example, 110v American outlet plugs are almost certainly not the most efficient way to power devices and appliances; but, because they are so standard and are good enough, using them massively eases the burden of making, buying and selling powered appliances in the US. Similar story with USB C.

> That’s not necessarily a bad thing.

True, and to my point, "popular" isn't necessarily "better".

> because they are so standard and are good enough, using them massively eases the burden of making, buying and selling powered appliances

Yes, but LLMs can do the heavy lifting when analyzing the "better programming languages" for productivity. There will be a bias toward popularity in their training sources, though.

Apparently, the Series library offers that. It didn't make it into the ANSI standard, but it's still maintained and covered in CLtL2.

edit SICP has examples on how to implement streaming (in Scheme).

From what I've seen, "AI" is typically written with the "Roman" (latin) letters, or translated as 人工知能 (AI) or as 生成AI (generative AI like LLMs).

Wikipedia is right; the romanization is just matching how it's rendered in kana.

I think your post pretty well illustrates how LLMs can and can't work. Favoriting this so I can point people to it in the future. I see so many extreme opinions on it like from how LLM is basically AGI to how it's "total garbage" but this is a good, balanced - and concise! - overview.

One approach would be to write code first, then run it by AI to get a critique. I think that strikes a good balance between avoiding atrophy and still getting the benefits of the tool.

If I create an executable with SBCL's save-lisp-and-die then run that executable, it presents a repl. Not really a development scenario, but more of a "running application" scenario. But, even then, if the executable creates a swank server, I can connect to that from SLIME.

When I saw the headline, I thought it referred to "Q" the array programming language (looks like your first link).

I didn't know about M-? ... I like how if there's only 1 match it automatically jumps to that single reference. :)