"Reducing typing" seems to be an important value for some software developers.
-
@simulo In general, my typing speed is rarely the bottleneck of my dev. Perhaps that's because I type at a pretty reasonable pace, and I use languages with very low amounts of boilerplate (and where nearly any boilerpoint can be abstracted away) and in the rare places where there are boilerplate (srfi-9 records come to mind) I have editor support to make typing them in fast through yasnippet.
In general, typing speed tends to correspond to the rate at which I can think, and a lot of the pauses come from needing to get up, pace around, think about the problem, get some tea, sit down again after a revelation. Or playing around with the idea iteratively and discovering the solution.
Lots of people are talking about LLMs as typing assistants and I just generally don't feel like that's a thing I need (excepting the RSI aspects; I do think more voice-driven editor piloting would be a good option to use sometimes)
-
@simulo To put it another way: a lot of people are now using tools that help them type *faster* than they can think. What are the consequences of that upon your work, and your understanding thereof?
I have also been talking to people who are like "yeah but AI is GREAT at refactoring"
I always say, refactoring is a really dangerous game, how often do you want to do it?
And then they say, I am just renaming symbols and concepts
Y'all I use emacs and grep and can rename the symbols and concepts of my codebase in minutes, I have never ever needed AI tooling to accomplish that. You're gonna need a different use case to sell me on
-
@simulo To put it another way: a lot of people are now using tools that help them type *faster* than they can think. What are the consequences of that upon your work, and your understanding thereof?
-
@simulo In general, my typing speed is rarely the bottleneck of my dev. Perhaps that's because I type at a pretty reasonable pace, and I use languages with very low amounts of boilerplate (and where nearly any boilerpoint can be abstracted away) and in the rare places where there are boilerplate (srfi-9 records come to mind) I have editor support to make typing them in fast through yasnippet.
In general, typing speed tends to correspond to the rate at which I can think, and a lot of the pauses come from needing to get up, pace around, think about the problem, get some tea, sit down again after a revelation. Or playing around with the idea iteratively and discovering the solution.
Lots of people are talking about LLMs as typing assistants and I just generally don't feel like that's a thing I need (excepting the RSI aspects; I do think more voice-driven editor piloting would be a good option to use sometimes)
Though lots of other lisp weenies seem to like LLMs. I do find I can just express what I want more clearly in a lisp (mostly clojure for me) than in English.
Again, as with you the few things like renaming or extracting stuff isn't something I'd need a LLM for, and I'd want deterministic results there
-
@simulo To put it another way: a lot of people are now using tools that help them type *faster* than they can think. What are the consequences of that upon your work, and your understanding thereof?
-
-
I have also been talking to people who are like "yeah but AI is GREAT at refactoring"
I always say, refactoring is a really dangerous game, how often do you want to do it?
And then they say, I am just renaming symbols and concepts
Y'all I use emacs and grep and can rename the symbols and concepts of my codebase in minutes, I have never ever needed AI tooling to accomplish that. You're gonna need a different use case to sell me on
@cwebber as a primarily C# programmer eveything you're saying here is how I'm feeling. I don't have any usecase where the existing tooling doesn't suffice! I am a little miffed at how Intellisense is less deterministic now and I'd like to look into how they might have ruined that in the past few years with LLM things, but I don't need anything to supplement the existing tooling.
The only thing that it seems to add onto the existing toolset is bad prototyping and bad test generation. I can do both of those myself and I'll learn more that way, so no thanks. -
R relay@relay.infosec.exchange shared this topic
-
"Reducing typing" seems to be an important value for some software developers. Is there some material origin to this? I.e. was too much typing ever a problem itself?
(Triggered by an add for an LLM that announced that programmers need to type less code, but the idea is not new)@simulo I've had co-workers who have trouble with their typing speed, but it has only ever affected prose: comments, documentation, emails, chat. The longer the text, the more they resists writing it. Code, however, has not been a problem to type for the people I've known.
-
-
"Reducing typing" seems to be an important value for some software developers. Is there some material origin to this? I.e. was too much typing ever a problem itself?
(Triggered by an add for an LLM that announced that programmers need to type less code, but the idea is not new)@simulo Most tooling just gets in the way. I need my editor to record what I type. That is all. It does not need an internet connection to interpret what I typed. Just accept what I typed already, miskates and all.
-
I have also been talking to people who are like "yeah but AI is GREAT at refactoring"
I always say, refactoring is a really dangerous game, how often do you want to do it?
And then they say, I am just renaming symbols and concepts
Y'all I use emacs and grep and can rename the symbols and concepts of my codebase in minutes, I have never ever needed AI tooling to accomplish that. You're gonna need a different use case to sell me on
@cwebber And whenever I do anything that changes a lot of code quickly, I always have to stop and think: was that a good idea? So even if it speeds up the typing, it doesn't speed up the analysis very much.
-
I have also been talking to people who are like "yeah but AI is GREAT at refactoring"
I always say, refactoring is a really dangerous game, how often do you want to do it?
And then they say, I am just renaming symbols and concepts
Y'all I use emacs and grep and can rename the symbols and concepts of my codebase in minutes, I have never ever needed AI tooling to accomplish that. You're gonna need a different use case to sell me on
@cwebber Yeah, like, IntelliJ has had _extremely_ powerful refactoring tools for decades that have not relied on generative AI in the slightest.
They aren't perfect, but they are a far cry better than the equivalents I've seen and used for generative AI.
-
"Reducing typing" seems to be an important value for some software developers. Is there some material origin to this? I.e. was too much typing ever a problem itself?
(Triggered by an add for an LLM that announced that programmers need to type less code, but the idea is not new)> Is there some material origin to this?
Like "Punchcard creation was really ehausting, thus clerks tried to reduce keystrokes" or the like.
-
@cmthiede @cwebber @simulo This.
The frustrating, time-consuming parts of the job tend to be the ones that are most rewarding and useful in the long term. They build my understanding or capabilities or expertise or empathy. They give me something to look back on with a sense of accomplishment. So few things in my life have been wasted effort. I grow; I develop.
Sure, automate the mechanical stuff that can be automated without loss of quality, stuff that takes time but doesn't require thought. Use grep and emacs to rename things, set up automated processes that take data from one form and populate the corresponding fields of a new form. Transcribe content from spoken to written so people who can't watch videos without getting headaches can read better transcripts.
But the keys are "automate" and "without loss of quality".
If all you've done is press an "easy" button, you haven't (and aren't) accomplished.
-
R relay@relay.mycrowd.ca shared this topic
-
"Reducing typing" seems to be an important value for some software developers. Is there some material origin to this? I.e. was too much typing ever a problem itself?
(Triggered by an add for an LLM that announced that programmers need to type less code, but the idea is not new)@simulo You've never seen a Java codebase with millions of lines of useless boilerplate? There's a manager class that believes this is what programming is, because LoC is their productivity metric and the way they justify pricing for contracted projects.
-
@simulo I have always assumed this is a sort of over-physicalisation from observing software devs. In other words, somebody watches you and observes that the only physical activity is typing, and therefore assumes that the entire activity is just typing. Not very different from when new students try to learn to write or read very fast. I think most experienced people would report that thinking is the bottleneck (in software dev and in studying).
@danielittlewood @simulo The only reason I learned to write fast is because a high school professor had all his notes pre-written on an overhead projector and I couldn't keep up.
It was useful later when I had to transcribe important information from talks at meetings for later prep into reports. For three years I spent about roughly two weeks every month generating 25 pages a day of handwritten content with figures, arrows, and direct quotes.
(Now I have arthritis and can't sign christmas cards without being in pain.)
this.
