> If you're reading this you probably have some idea of what slate is.
Tip for anyone writing technical/product blogs: never assume readers know what your product is. They probably came in for the first time from a random link on some site like HN and this is your first (and maybe only) opportunity to get that elevator pitch in. I think this particular example is a good one, in that they immediately provide a one-sentence "refresher" that gives the rest of the post some context.
Might be also worthwhile considering a more memorable/distinctive name, unless of course this is only aimed at those who are actively and aggressively following developments in that space. The post piqued my interest enough to make me vaguely think, hmmm, I might follow that up some time in the future when I'm less busy however if I fail to bookmark it and assuming the name 'slate' loosely sticks in my head, then a Google search for 'slate', 'slate ai' or even 'slate ai agent' isn't going to bring me back to this product again.
I cannot find any link to a repository with that ported library. I can't imagine why you wouldn't share this exciting port, unless of course the code is not that great?
And as a sidenote, why would you ever hijack the up/down cursor keys on a web page? There are still people out there on actual computers which use them for scrolling.
> And as a sidenote, why would you ever hijack the up/down cursor keys on a web page? There are still people out there on actual computers which use them for scrolling.
There are also people who will give up on the article when they realize they can't scroll with the arrow keys.
How would you verify that the ported code actually works if you don't port the tests and examples?
I see tremendous use in a tool that could be used to port "any" library to any language. I'm very skeptical if this works if the library itself depends on other language specific libraries, ... but we'll see.
If you'd seriously want to use that port in production it only makes sense to port the tests and examples. How would you verify the ported code works otherwise? This also means that if the original library has bad test coverage.. you have too.
It's much more convincing evidence that the port worked if you run the same tests and examples as before. If you've let the llm change the tests too it's very hard to tell whether the new codebase does the same thing as the original.
Ah yes, reading fail on my part. Yes, successfully compiling python to typescript would be a headache. I think some things would essentially have to turn into eval of strings. Though not if the project was sufficiently uninteresting python.
> If you're reading this you probably have some idea of what slate is.
Tip for anyone writing technical/product blogs: never assume readers know what your product is. They probably came in for the first time from a random link on some site like HN and this is your first (and maybe only) opportunity to get that elevator pitch in. I think this particular example is a good one, in that they immediately provide a one-sentence "refresher" that gives the rest of the post some context.
Might be also worthwhile considering a more memorable/distinctive name, unless of course this is only aimed at those who are actively and aggressively following developments in that space. The post piqued my interest enough to make me vaguely think, hmmm, I might follow that up some time in the future when I'm less busy however if I fail to bookmark it and assuming the name 'slate' loosely sticks in my head, then a Google search for 'slate', 'slate ai' or even 'slate ai agent' isn't going to bring me back to this product again.
> If you're reading this you probably have some idea of what slate is.
Yeah, they make audio plugins. :)
It's a type of rock
I cannot find any link to a repository with that ported library. I can't imagine why you wouldn't share this exciting port, unless of course the code is not that great?
And as a sidenote, why would you ever hijack the up/down cursor keys on a web page? There are still people out there on actual computers which use them for scrolling.
> And as a sidenote, why would you ever hijack the up/down cursor keys on a web page? There are still people out there on actual computers which use them for scrolling.
There are also people who will give up on the article when they realize they can't scroll with the arrow keys.
I was looking for signs that the tests and examples worked afterwards. Those are missing.
Its javascript to typescript, which unless I'm missing something, can be done with `mv foo.js foo.ts`. No examples of any transforms done by the tool.
And it wasn't one sentence, it was a sequence of questions and answers.
I guess they started with the title for the blog post and didn't really worry about showing whether it did anything useful or not.
> .. excluding all tests and examples, ..
How would you verify that the ported code actually works if you don't port the tests and examples?
I see tremendous use in a tool that could be used to port "any" library to any language. I'm very skeptical if this works if the library itself depends on other language specific libraries, ... but we'll see.
If you'd seriously want to use that port in production it only makes sense to port the tests and examples. How would you verify the ported code works otherwise? This also means that if the original library has bad test coverage.. you have too.
It's much more convincing evidence that the port worked if you run the same tests and examples as before. If you've let the llm change the tests too it's very hard to tell whether the new codebase does the same thing as the original.
Yes, that would be even better - but also way harder for unit tests as you’d need to interop between the langauges.
Claude Sonnet has a funny habit of fixing test failures by deleting the failing tests.
It really is like an over enthusiastic college intern!
> How would you verify that the ported code actually works if you don't port the tests and examples?
Port tests, run them against unported code. Port code, run it against unported tests? That way you also verify that the public api hasn't changed
it was Python -> Typescript
Python to typescript is far more impressive. Assuming it's working of course.
Ah yes, reading fail on my part. Yes, successfully compiling python to typescript would be a headache. I think some things would essentially have to turn into eval of strings. Though not if the project was sufficiently uninteresting python.
Why trumpet code that is so ready for the garbage that you wouldn't even bother to publish it
seriously. this is no different from that kid who ran Claude code in a while loop
Essentially this is Ralph as a service - https://ghuntley.com/ralph
The post’s title is mostly occluded by an image on mobile, so that bodes well.
What is slate? any link to it?