Colloidal@programming.dev
on 26 Jul 18:38
nextcollapse
“So instead of writing three applications, you write it in a special programming language, which is basically English, which describes how you want to see this application in a very specified way, and then AI agents, together with JetBrains tooling, will generate the code of all of these platforms,” Skrygan said.
This makes me think of Inform, which compiles English sentences into interactive fiction.
tavernusmaximus@piefed.social
on 26 Jul 18:50
nextcollapse
The language is in the works but JetBrains has not revealed a timeline for general availability at this point.
Won’t hold my breath for this ever shipping.
Assuming there will be an LLM involved because that’s what seems to be all AI is these days. How on earth they plan to get reproducible builds from this thing is beyond me (suppose that’s one reason I don’t work for JetBrains).
nebeker@programming.dev
on 26 Jul 19:47
nextcollapse
Surely through an intermediate - real - language?
BlameTheAntifa@lemmy.world
on 27 Jul 00:16
collapse
It sounds like it uses similar ideas to Amazon Kiro. Many of the advancements in “vibe coding” tools are focused on ways to put consistent, coherent bumpers on AI output.
jubilationtcornpone@sh.itjust.works
on 26 Jul 18:54
nextcollapse
This kind of seems like a solution in search of a problem. Most modern high level programming languages are easily readable, ‘english oriented’, and already capable of at least some level of cross platform development.
One of the main problems with any programing language or framework is that flexibility breeds complexity. If they seriously think they’re going to lower the complexity of programming by allowing devs to write programs [essentially] in plain English, and then let AI do the rest, I think it’s a recipe for disappointment.
Michal@programming.dev
on 26 Jul 20:30
nextcollapse
This kind of seems like a solution in search of a problem
Not like it’s a bad things. A lot of inventions started this way.
Yes, and it’s good that programming languages are still experimented with, otherwise we’d still be writing assembly.
douglasg14b@lemmy.world
on 27 Jul 08:58
nextcollapse
Naw, This is honestly the direction that software engineering is going to go. AI becomes more capable over time.
We are eventually going to stop writing code and focus more on writing specifications. The development of languages that allow us to write and maintain better specifications is going to accelerate that in the same way, that higher level languages allowed us to accelerate writing code for the purpose of it being transformed into some form of bytecode. We are now in the early stages of needing a language that better facilitates the authoring of detailed specifications that can then be ran through code generation in more predictable and scalable manners.
I see nothing wrong with developing a new language. If it works it works. If it doesn’t it doesn’t and we all learned new shit. I’m not sure why so many people in this thread hate science.
jkercher@programming.dev
on 27 Jul 10:49
nextcollapse
We are eventually going to stop writing code and focus more on writing specifications.
I don’t think this will happen in my lifetime.
Swedneck@discuss.tchncs.de
on 28 Jul 21:19
collapse
i mean isn’t it a kind of old thing in that companies usually have somewhat separate designers and programmers?
I don’t understand why you’re getting downvoted. While I don’t share your conviction, I do admit it’s certainly a possibility.
The advantage of doing things that way is that code becomes much more portable. We may finally reach the goal of “write once, run anywhere”, because the AI may write all the platform specific code.
It does make a big assumption that the AI output is reliable enough though. At times people will want to tweak the output, so how are they gonna go about that? Maybe if the language is based on Markdown, you can inject snippets of code where necessary. But if you have to do that too often, such a language will lose its appeal.
There’s a lot of unknowns, but I see why it’s a tempting idea.
state_electrician@discuss.tchncs.de
on 27 Jul 11:22
nextcollapse
I don’t have high hopes for it to be actually useful. They already have Kotlin, which feels like a playground for language designers.
“JetBrains is exploring how to make this new language a derivative from Kotlin, but Skrygan believes the derivative should be English.”
That sounds like (Visual) Basic.
It looks like English but it’s basically pseudo-code.
I’m happy letting AI and my language server write all the extra annotations for Rust, i’ve no trouble reading them.
I have much more trouble when types and usage specifiers/limiters are missing.
resipsaloquitur@lemmy.world
on 26 Jul 22:29
nextcollapse
I read that part. It’s just that isn’t really a proposal on “how” and more a tagline of every high level programming language since the 80s, including the AI all-star, Python
AnotherPenguin@programming.dev
on 27 Jul 01:27
nextcollapse
Do they really need it when kotlin exists?
LeFantome@programming.dev
on 28 Jul 01:37
collapse
What they are aiming for (not agreeing, just explaining) is a language that you can use to ask AI to do things for you.
The idea is that you do not have to do the nuts and bolts programming (the AI will do that) but at the same time you have more deterministic control over what the AI does.
So “higher level” than our highest level languages now.
6nk06@sh.itjust.works
on 27 Jul 06:44
nextcollapse
which is basically English, which describes how you want to see
It will be a mix between Basic and Cobol. I like what JetBrains do, but that’s a stupid idea.
abbadon420@sh.itjust.works
on 27 Jul 07:40
collapse
I’m guessing it’ll be closer to bytecode, since they’re very java focused.
abbadon420@sh.itjust.works
on 27 Jul 12:20
collapse
That is not what I said.
They’re designing a language that supposedly writes like English. Why would you not compile English straight to bytecode? Since they made Kotlin run in the jvm, they’re likely gonna make this run on the jvm as well. They could make another intermediate language, but they could also make an advances interpreter or compiler that compiles english straight to bytecode. Nobody needs to read or write that, they reading and wrting English.
JakenVeina@midwest.social
on 27 Jul 07:54
nextcollapse
you write it in a special programming language, which is basically English, which describes how you want to see this application in a very specified way
A derivative of English, with different syntax and rules to help eliminate ambiguities? We in the industry tend to call that “code”.
That’s a long list of words I have never heard of. I do like how SQL somehow fits the bill of a 4GL.
hornywarthogfart@sh.itjust.works
on 28 Jul 00:21
collapse
It makes sense, you aren’t telling sql server how to do something, you just tell it what you want and it figures it out. You aren’t even doing procedural stuff at that point.
I like the RAD tools being qualified as 4GLs as I haven’t really thought of them that way but again it makes sense.
Also screw PowerBuilder. I am sorry if anyone in this thread likes it…but it is seriously awful.
Edit: Before people jump me, I do know that you have some influence over execution plans with join orders, hints, etc… but by and large you don’t tell SQL Server how to do it’s job.
threaded - newest
This makes me think of Inform, which compiles English sentences into interactive fiction.
Won’t hold my breath for this ever shipping.
Assuming there will be an LLM involved because that’s what seems to be all AI is these days. How on earth they plan to get reproducible builds from this thing is beyond me (suppose that’s one reason I don’t work for JetBrains).
Surely through an intermediate - real - language?
It sounds like it uses similar ideas to Amazon Kiro. Many of the advancements in “vibe coding” tools are focused on ways to put consistent, coherent bumpers on AI output.
This kind of seems like a solution in search of a problem. Most modern high level programming languages are easily readable, ‘english oriented’, and already capable of at least some level of cross platform development.
One of the main problems with any programing language or framework is that flexibility breeds complexity. If they seriously think they’re going to lower the complexity of programming by allowing devs to write programs [essentially] in plain English, and then let AI do the rest, I think it’s a recipe for disappointment.
Not like it’s a bad things. A lot of inventions started this way.
But also, a lot of programming languages exist simply because a programmer really wanted to write a programming language.
Yes, and it’s good that programming languages are still experimented with, otherwise we’d still be writing assembly.
Naw, This is honestly the direction that software engineering is going to go. AI becomes more capable over time.
We are eventually going to stop writing code and focus more on writing specifications. The development of languages that allow us to write and maintain better specifications is going to accelerate that in the same way, that higher level languages allowed us to accelerate writing code for the purpose of it being transformed into some form of bytecode. We are now in the early stages of needing a language that better facilitates the authoring of detailed specifications that can then be ran through code generation in more predictable and scalable manners.
I see nothing wrong with developing a new language. If it works it works. If it doesn’t it doesn’t and we all learned new shit. I’m not sure why so many people in this thread hate science.
I don’t think this will happen in my lifetime.
i mean isn’t it a kind of old thing in that companies usually have somewhat separate designers and programmers?
I don’t understand why you’re getting downvoted. While I don’t share your conviction, I do admit it’s certainly a possibility.
The advantage of doing things that way is that code becomes much more portable. We may finally reach the goal of “write once, run anywhere”, because the AI may write all the platform specific code.
It does make a big assumption that the AI output is reliable enough though. At times people will want to tweak the output, so how are they gonna go about that? Maybe if the language is based on Markdown, you can inject snippets of code where necessary. But if you have to do that too often, such a language will lose its appeal.
There’s a lot of unknowns, but I see why it’s a tempting idea.
I don’t have high hopes for it to be actually useful. They already have Kotlin, which feels like a playground for language designers.
Yeah… Kotlin is an unreadable nightmare.
It’s the “I’m scared of brackets” crowd again.
“JetBrains is exploring how to make this new language a derivative from Kotlin, but Skrygan believes the derivative should be English.”
That sounds like (Visual) Basic. It looks like English but it’s basically pseudo-code.
I’m happy letting AI and my language server write all the extra annotations for Rust, i’ve no trouble reading them. I have much more trouble when types and usage specifiers/limiters are missing.
Does it involve AI?
Yep!
Don’t worry, no details on how a language would achieve that
Oh okay.
The idea is that the programming language is closer to natural language, which is what LLMs do best.
I read that part. It’s just that isn’t really a proposal on “how” and more a tagline of every high level programming language since the 80s, including the AI all-star, Python
Do they really need it when kotlin exists?
What they are aiming for (not agreeing, just explaining) is a language that you can use to ask AI to do things for you.
The idea is that you do not have to do the nuts and bolts programming (the AI will do that) but at the same time you have more deterministic control over what the AI does.
So “higher level” than our highest level languages now.
It will be a mix between Basic and Cobol. I like what JetBrains do, but that’s a stupid idea.
I’m guessing it’ll be closer to bytecode, since they’re very java focused.
Nobody writes byte code directly.
That is not what I said.
They’re designing a language that supposedly writes like English. Why would you not compile English straight to bytecode? Since they made Kotlin run in the jvm, they’re likely gonna make this run on the jvm as well. They could make another intermediate language, but they could also make an advances interpreter or compiler that compiles english straight to bytecode. Nobody needs to read or write that, they reading and wrting English.
Also, yes they do. www.youtube.com/watch?v=GbNv6rdYJL0
Maybe not many and even less do it proffesionally, but its not none
A derivative of English, with different syntax and rules to help eliminate ambiguities? We in the industry tend to call that “code”.
This is how cobol was created
Another chatgpt wrapper sold as gold subscription.
That’s my bet.
So another 4th generation language that no one uses. Great.
That’s a long list of words I have never heard of. I do like how SQL somehow fits the bill of a 4GL.
It makes sense, you aren’t telling sql server how to do something, you just tell it what you want and it figures it out. You aren’t even doing procedural stuff at that point.
I like the RAD tools being qualified as 4GLs as I haven’t really thought of them that way but again it makes sense.
Also screw PowerBuilder. I am sorry if anyone in this thread likes it…but it is seriously awful.
Edit: Before people jump me, I do know that you have some influence over execution plans with join orders, hints, etc… but by and large you don’t tell SQL Server how to do it’s job.
Never used Power Builder, but used to touch Pro4 and Pega Systems in different jobs. They were both middle manager programming UI’s
They need to call it COBOL. A language regular business people can use!