Interesting whether guys employ LLM to speed up development. Starting a new browser just like this would be very bold decision like 15 years ago, now seems like a reasonable thing to do actually.
Would have to be at the hands of an already skilled practitioner. Average Joe programmer setting out to build a browser with Copilot will end badly. Big-picture architecture and discipline is too important with this level of complexity.
Not what I meant really. antirez is not a junior yet codes with LLM assistence.
Not sure why this gets downvoted but since when is it illegal to ask whether people actually use intellisense? I mean - what’s wrong with being interests how ppl bootstrap and get to thousands of LOCs in practically no time.
Starting a new browser, using LLMs... Is not going to maintain enough context.
Whilst Andreas does use Copilot a fair bit [0], he tends to do a line at a time, frequently disagree and rewrites his own, before prompting again. That is... He basically uses it as a fancy autocomplete. Not much else.
Fancy or not an autocomplete is autocomplete. What if it autocompletes one class or one function at a time? Does this make it obnoxious to even think asking about it on HN?
We've had macro expansion since the 80s. If your IDE can't autocomplete, then it's a text editor and not an IDE. Which makes it an irrelevant question to ask, because it has no impact on making something feasible or not.
Of course, brother, that’s the point. And while oldies one by one get to pair their IDEs with RAG+LLM autocompletes, some brothers keep raging against people using higher level code assistance.
I haven't watched the VODs myself but I am curious as to the nature of the Copilot use - is it a fancy autocomplete, a 2nd look at code he's already written, or is it generating large parts of the code?
The SerenityOS project literally has "NIH syndrome" at its core ["the SerenityOS project, which has a strong culture of writing everything from scratch.", https://ladybird.org/#faq] so I would expect a degree of skepticism of AI code generation.
But there was the post semi-recently about Cloudflare developers shipping a mostly-AI-generated software, overseen by real devs [https://news.ycombinator.com/item?id=44159166] Perhaps Andreas' use of Copilot is a similar thing, where he saves keystrokes and finds errors in the generated output with his own experience. I'll go watch some VODs.
Most of it was written without LLMs for that reason, but now Ladybird is its own project, the NIH focus has largely gone away. They are using a lot of third party code now. Their mission is compete with Chrome, not be 100% NIH.
The original project did everything from scratch. Video/audio decoding, SSL, font rendering, etc. a huge portion of the code.
They're still not using existing browser code because being an independent implementation of the web standards is a key goal. They've found bugs in the standard just by implementing it.
A comment that could apply to anything. Notice how nobody except for your subthread here is talking about AI. This thread is not about AI or LLMs, for a good reason.
they’re downvoting you, but the developer actually uses copilot a lot in his development videos. why are people so up in arms?
of course, he’s a very, very proficient developer and a browser specialist. he’s not just vibecoding, like you might be implying. but he also uses llms for development.
Because it's bad enough (for those of us that aren't heavily invested in the topic) that often 50%+ of the frontpage is about AI now, to go into the comments for a clearly non-AI topic and see a bunch of "but what about LLMs" comments really just sucks the fun out of HN for a lot of us.
That’s not even close to true unless you rate all ML as AI and all conditional structures thinking. Like, why don’t you first try to rate and count them before spitting malice ?
in all honestly much less AI stuff gets to the HN front compared to when OpenAI first released GPTs. Given there are thousands of publications every month and dozens of startups launched with ridiculous valuations and most of this is actually submitted here for karma or leverage.
Besides, what, knowledge representation via stochastic probabilistic graphical models is now an heresy? Since when? Since MPEG or MP3 ?
Well to my surprise they do, even though every sane dev in 2025 uses some sort of code assîtes autocomplete. So it seams now we discussing what is the threshold reg intelligence of the autocomplete.
Ridiculous stuff indeed.
I’ve been coding for 25 years and nothing in my message implies vibe coding. I know what I’m talking and sure bet many others here do.
... and if you (the parent comment author) wants a really easy tool for creating a UI for the Python code, I'd recommend looking at Streamlit: https://streamlit.io/
I'd go out on a limb and say the rock-solid out-of-the box experience is what is keeping many people using R-studio (and even R itself), rather than the messy ecosystem of Python. I'm seeing this tendency in myself for some tasks.
Also, my impression is that that is also a big part of why MATLAB still exists, despite outraging prices.
I think the common theme among these tools' main user groups is that they are not developers. They are not comfortable fiddling a lot with a dev environment, but can be productive in an environment where everything just works.
Thus, if Positron can get the same smooth and rock-solid out-of-the box experience, it will be able to reach a lot of these non-developer user groups.
Jup, that's 100% it. Even tho I build productive systems in anything but R, whipping up Rstudio and having it all included or easily installable without any package manager fuckery or anything makes it a non brainer.
To implement an efficient dataflow-based programming API/DSL, you better have some support for channels and lightweight threads in a scriptable language, something that you've got in Groovy with the GPars library that Nextflow uses.
We opted for implementing all of this in Go in SciPipe, where we get similar basic dataflow/flow-based functionality as Nextflow with the native concurrency primitives of Go, but the Go syntax probably/surely puts away some biologists who have written some python at most before, and Go won't let us customize the API and hide away as much of the plumbing under nice syntax, as Groovy.
In this regard, Groovy with the GPars library for the concurrency, doesn't seem as a particularly bad choice. There weren't that many options at the time either.
The downside has been tooling support though, such as editor intelligence and debugging support, although parts of that is finally improving now with a NF language server.
Today, one could probably implement something similar with Python's asyncio and queues for the channel semantics, and there is even the Crystal language that has Go-like concurrency in a much more script-like language (see a comparison between Go and Crystal concurrency syntax at [1]), but Crystal would of course be an even more fringe langauge than Groovy.
The 99% identity with chimps is extremely misleading without clear qualification.
The fact that it has been used without qualification has a lot to do with the fact that most of our genome was assumed to be junk, which we know today is not the case, per the ENCODE project for example.
Thus, the 99% number needs to stop being perpetuated.
Today we know that the alignable parts (parts that are similar enough that they can be aligned with each other) are down in the 80s percents between humans and chimps, which can be digged out from e.g. this recent big study (some numbers are in the paper, and some needs to be digged out of the supplementary material):
In the early 2000s, XSLT allowed me as a late teenager with some HTML experience but without real coding skills (I could copy some lines of PHP from various forums and get it to work) to build a somewhat fancy intranet for a local car shop, complete with automatic styling of a feed of car info from a nationwide online sales portal.
Somehow it took me many years, basically until starting uni and taking a proper programming class, before I started feeling like I could realize my ideas in a normal programming language.
XSLT was a kind of tech that allowed a non-coder like me to step by step figure out how to get things to show on the screen.
I think XSLT really has some strong points, in this regard at least.
And it doesn't hurt that Andreas seems to be such a nice, humble guy.