Time Nick Message 11:55 rubenwardy do any languages translate proper names like macOS, Linux, FreeBSD? 11:59 celeron55 that seems like the perfect question for an LLM 12:00 celeron55 it seems translation is super rare but transliteration is more common, for other alphabets 12:01 sfan5 in japanese you could replace Windows with ウインドウズ, but I don't think that's a common thing to do 12:01 rubenwardy the reason I ask is I didn't make these translatable on the website originally, but sfan5's PR has added them to the translation template 12:02 celeron55 "Languages using Cyrillic, Greek, Armenian, or Georgian scripts usually transliterate rather than translate" 12:02 celeron55 "Arabic, Hebrew, Persian, Urdu, etc. also just transliterate into their own script: لينكس / Linux, ماك أو إس / macOS, فري بي إس دي / FreeBSD" 12:02 celeron55 i don't know if this is true or not. so: AI slop warning 12:03 sfan5 rubenwardy: it was inconsistent before, 'Windows' was already being passed through i18n 12:12 sfan5 how does the language redirect work anyway? 12:13 sfan5 I have Accept-Language "de-DE,en-US;q=0.7,en;q=0.3" and I get the english website version 12:17 rubenwardy it's currently a fixed redirect to the english version 12:17 rubenwardy there's an issue to change this 12:17 sfan5 oh huh 12:17 rubenwardy web#365 12:17 ShadowBot https://github.com/luanti-org/luanti-org.github.io/issues/365 -- Support Accept-Language 12:19 rubenwardy One thing that makes this less bad is that search engines will show you your language in the results 12:19 rubenwardy like if you google Luanti in german, it should show /de/ 12:19 rubenwardy once google realises 12:20 sfan5 it indeed does 14:08 MTDiscord I think there is a low hanging fruit of writing a clean wrapper API around our current HUD API 14:09 MTDiscord I'm considering first exploring this as a mod, and maybe if that is received well, upstreaming it 14:09 MTDiscord glances at modlib 14:10 MTDiscord don't worry, this will have a more well defined purpose 😄 14:10 MTDiscord I also have a branch lying around somewhere where I made sure that the C++ only gave you unique IDs, but I figured it's likely cleaner to just use an OO Lua wrapper instead 14:11 MTDiscord The reason I started thinking about this again is because I saw some of Zughy's PRs related to HUD things, and even simple things suffer greatly from the fact that we don't abstract over the gory "everything is squeezed into the same struct" 14:12 MTDiscord real question tho, is it really worth time on the current hud/formspec mess or just spending some time on vrobs new unified gui or something? 14:14 MTDiscord basically what your proposing (assuming actual adoption/being recieved) is going to take at least 3-4 release minium, at which point.... (yes i know this is the engine that takes forever) 15:11 [MatrxMT] v-rob: if you can read us, what's a realistic status of your GUI/HUD overhaul? 15:12 [MatrxMT] The faster I start using something better than we currently have, the better 15:12 [MatrxMT] *than what we 15:14 MTDiscord wsor: that's exactly what this is about :) i think that with modest effort, i can write a modder-friendly wrapper that also rules out some kinds of bugs. 15:15 MTDiscord whether it'll be worth upstreaming.. well, we'll see. 16:31 MTDiscord My backup plan for if vrob never finishes is waiting for LLMs to improve even further and then have it help me get that PR over the finish line 16:47 rubenwardy friends don't let friends vibe code a GUI API 16:53 celeron55 LLMs already are an incredible help in any coding task but you do need to know quite specifically what you want. you still can't just unleash them and expect it to make reasonable architectural choices on its own 16:53 MTDiscord Well aware, just my internal plan, not like it wouldn't get judged by everyone else and thoroughly tested 16:54 sfan5 I can think of many more critical areas to vibe code 16:54 sfan5 not that that makes a good idea 16:54 MTDiscord Haha, anyways what I got out of this is rubenwardy might be my friend 16:55 celeron55 i vibe code almost every day at work. ask me anything about it 16:58 MTDiscord What stack, and level of complexity? Is this a database with a front end and an API? A car simulation? Medical equipment firmware? 16:59 rubenwardy it's cloudflare middleware 17:01 celeron55 i do embedded programming and related tools. the program scope being limited by hardware limitations makes it easy to manage context 17:01 celeron55 embedded as in not even linux, bare hardware 17:02 sfan5 >innerHTML = "€ " + (member.stats.totalDonations / 100).toFixed(0); 17:02 sfan5 who comitted this sin? 17:02 sfan5 the euro symbol goes after the number 17:03 sfan5 nvm apparently english does it like that >.> 17:06 celeron55 (oh and: mostly rust and python) 17:13 MTDiscord Gotcha, yep. Honestly there are few production code bases I've seen that enterprise sized LLMs fail entirely at. Obviously they aren't perfect, but had a lot of success in leveraging them as a tool 17:14 rubenwardy sfan5: I was just copying the format given on "Top financial contributors" on https://opencollective.com/Luanti 17:14 rubenwardy but can move it, doesn't really matter 17:14 rubenwardy maybe there's a locale way of rendering it 17:15 sfan5 somehow probably 17:15 sfan5 that page shows "5 € EUR / month" (why both?) in the plans for me but the backer list is still "€25 EUR since XXX" 17:20 rubenwardy new Intl.NumberFormat("de-DE", { style: "currency", currency: "EUR" }).format(number) 17:21 rubenwardy can read locale from the html element 17:21 [MatrxMT] As far as I know, we put the currency first in Italian as well 17:22 celeron55 @exe_virus i think one can think of LLMs for coding a bit like a concrete pour for construction. As long as you set up the mold, the concrete will do the bulk of the work very fast 17:22 MTDiscord Good analogy 17:26 celeron55 you can't make artsy brickwork with it, but usually that's not specfied either, for something that's only specified to be functional 17:46 sfan5 should we replace the main, recommend download link with a button? 17:53 sfan5 https://0x0.st/KCiQ.png example 17:56 celeron55 the common person is probably more likely to click a button rather than a link. that button text isn't the most palatable though, if that's the point 17:57 celeron55 i guess it's a balance between trying to keep the user informed while already knowing what they want and only really needing to get in a click? 17:59 celeron55 a company trying to sell its product would probably launch a modal dialog "don't waste more time, click here, we know you want to" 18:00 sfan5 18:04 sfan5 i think it's a clear improvement 18:04 sfan5 it only gets hairy for macOS because we have 4 "recommend" downloads there 18:06 celeron55 i agree it's better than the current page. i'm just unsure what the actual peak optimal download link design would be 18:06 sfan5 to be clear with https://github.com/luanti-org/luanti-org.github.io/pull/399 I expect the average user to see only exactly one download button on his screen 18:08 sfan5 and right in the middle too 18:27 MTDiscord re "vibe coding", what i'm currently wondering about is using LLMs to assist turning lua_api.md (an informal specification in prose) into a "formal" specification (e.g.: at least some basic types) that we can then turn into a format tools like LuaLS can work with 18:34 MTDiscord that website pr seem suboptimal if thats the case (only one download). any sane website lets you at least pick another platform if not having them listed 18:34 sfan5 the other sections are hidden behind a click 18:35 MTDiscord ah, yeah i see that in your js 18:35 MTDiscord js be looking like python lol 19:25 sfan5 planning to merge #16728, #16727, #16732 in 15m 19:25 ShadowBot https://github.com/luanti-org/luanti/issues/16728 -- Fix errors with fs::RecursiveDelete() on paths that don't exist by sfan5 19:25 ShadowBot https://github.com/luanti-org/luanti/issues/16727 -- Remove hardcoded color of infobox in /help by Wuzzy2 19:25 ShadowBot https://github.com/luanti-org/luanti/issues/16732 -- Breakages: split fog from sky functions by Zughy 20:30 MTDiscord Absolutely I'd use LLMs to assist with the conversion of lua_api.md, but still need to drill down on a format for documentation we like before you can ask a model to do the changes 22:49 MTDiscord Im curious if the stance on AI assistance for a PR has changed: If I were to contribute to the engine I would want to use AI to some extent. Not to prompt it to make desired feature, but to write snippets that I incrementally build up into the pr and test as I go. Is that type of contribution / ai use frowned upon still? 23:05 rubenwardy > Looking for older versions? Check out the list of releases on GitHub! 23:05 rubenwardy I feel like I'd be looking for "Show all versions" here 23:05 rubenwardy *all platforms 23:05 rubenwardy but nice to have this