I've had it write Scheme with little issue -- it even completely the latter half of a small toy compiler. I think the REPL is the issue, not the coding; forcing it to treat the REPL like another conversation participant is likely the only way for that to work, and this article does not handle it that way. Instead, hand it a compiler and let it use the workflow it is optimized for.
Agreed. The article bemoans the fact that AIs don’t need to work in the inefficient way that most humans prefer, getting micro-level feedback from IDEs and REPLs to reduce our mistake count as we go.
If you take a hard look at that workflow, it implies a high degree of incompetence on the part of humans: the reason we generally don’t write thousands of lines without any automated feedback is because our mistake rate is too high.
I will guess that farm subsidies go to... well, farming. Doesn't sound completely ridiculous at first sight.
> you're mad about space exploration?
Exploration? It's not exploration at all: it's sending 4 humans for a 10 days trip around the Moon. I wish they used the money for actual space exploration, though.
Hooking up to and generating calls across filesystem APIs cost multiple orders of magnitude more than calling `ls`. These tooling ideas are interesting, though. Maybe Kenneth_E._Iverson was right all along?
Talking to another senior dev over drinks tonight, we both worried not about our work but about who might come up never having written a single line of code. Never even opened a terminal. Is looking at the code something you learn in semester 5?
I think computer science education is going to stomp onward, poorly. And we will get that generation. And things like "terminal tooling is going out of style" won't even be said any more. Hacker groups will turn from discussions about new ideas to talking about doing leetcode without AI.
Our art died because we used our art to kill it. We are the last human masters.
>Hooking up to and generating calls across filesystem APIs cost multiple orders of magnitude more than calling `ls`. These tooling ideas are interesting, though. Maybe Kenneth_E._Iverson was right all along?
I don't think you understood what the idea was. Its not about calling `ls` vs not. I don't think UNIX commands are going away (or at least deterministic calls).
Its the interface itself that would go away. We won't work on terminals but some other interface which would use commands internally.
It's definitely _alluded_ to in Player of Games, mostly as a method of emphasising how unpleasant the Azad society is, but I don't think we ever really _see_ any of it?
(I may just be forgetting; it's probably at least a decade since I last read it.)
If I remember correctly, Gurgeh's internal narrative reveals what he observes in one of the secret channels that are restricted to the upper class. Unpleasant is quite the understatement. lol
But yeah, I don't think we actually see it firsthand, it's more disturbing than gory.
As someone who has implemented full match in several industrial languages: this isn’t really match; it doesn’t not handle unpacking. And that is by far the only interesting bit. This feature is more accurately called `cond` à la Scheme, and you can fully expand it away ahead of type checking. Looking at unpacking in the arms, even with Scheme’s truth-y values and `=>`, could be neat.
Optimizing well-known jumps is useful, as is branch reordering, but the tombstone flag is unnecessary; you can simply write down a list of all targeted / called blocks and perform dead block elimination more generally that way.
Look at the compaction prompt yourself. It's in my opinion way too short. (I'm running on Opus 4.5 most of the time at work)
From what my colleague explained to me and I haven't 100% verified it myself is that the beginning and end of the window is the most important to the compaction summary so a lot of the finer details and debugging that will slow down the next session get dropped.
reply