Community
Chat Logs
Tuesday, October 20, 2026 (America/New_York)
- bluehype https://repopack.com/project/rcompat/tasks/61
- dreamrealhttps://bytecode.news/posts/2026/10/there-is-no-irrelephant
- bluedreamreal: wow
- dreamrealtold you I thought it was a firecracker
- bluedamn
- dreamrealAs always: if you think the content's valuable, or might be valuable to someone, please share it: without grassroots involvement, that site runs on whatever passion I have room to give it, and I'm *busy*
- blueI just realised that since I restart the server, I'd only rejoined #primate...
- blueI should add everything to my autojoin, but can't be bothered
- bluerestarted*
- dreamrealwhich client do you use?
- blueirssi
- dreamreal*nod* I don't remember how to configure it in irssi, I moved to weechat a long time ago and don't remember why
- bluedreamreal: I *love* compaction in repopack. it's basically like to distilling everything that happened before, and creating a new baseline for tasks. It's amazing because it's a much better way to move tasks with changing requirements forward, than editing notes
- bluehttps://repopack.com/project/repopack/tasks/283
- blueyou probably can't see that... but it looks REALLY nice
- blueI tried to give you read access to tasks on repopack, but I think you never accepted the invite
- dreamrealI don't remember that, either - I might not have, I get a ton of data across all input streams and I don't want to use tooling to sort it because I'm human, damn it!
- blueoh well, if you're ever interested or have the time, drop me a message
- dreamrealYou saw the issue I filed sunday or monday, yeah?
- bluehttps://repopack.com/project/primate/tasks/231
- bluethis?
- blueno I didn't, I was off monday
- bot[ByteCode.News] There Is No Irrelephant - https://bytecode.news/posts/2026/10/there-is-no-irrelephant
- dreamrealIs that the <head> bug?
- blueyes
- dreamrealthen yeah.
- dreamrealIt's not a huge thing, just duplicated data
- dreamrealand I think the OG stuff that requires node_modules is because the modules I used kinda suck
- dreamrealthey work but meh
- bluewell yeah, need to get rid of that
- bluenode_modules has lost nothing in deployment
- dreamrealWorking on backend streampack issues first, for forge subscriptions
- blueyou don't really put in your entire maven tree in java deployment either, as far as I know
- dreamrealwell, I think the problem is that they look stuff up at runtime
- dreamrealbut I'm not sure, I just needed OG to work pronto
- blueprimate actually copies node: addons too into the build folder. the idea is that the build folder must be self contained. you should be able to copy it whole and deploy from anywhere
- blueif this invariant is violated, it can have two reasons. either you messed up, or primate messed up
- blueI believe there was a marginal case where primate could not copy the .node addons due to some weird dynamic shenanigans
- blueor perhaps the mechanism just doesn't work, I haven't used it myself often, it worked the last time I did
- blueanyway, if it's an in-tree issue, I suggest you fix it, it'll improve the security and speed of deployment. if it's a primate issue, please file a task
- dreamrealI don't know, will have to look - it was satori, I think, if memory serves - I was *busy*
- dreamrealand was like "oh crap, why is OG failing, must make it work"
- blueopengraph?
- bluehttps://www.npmjs.com/package/satori ?
- bluehttps://repopack.com/project/primate/tasks/231 is done
- dreamrealI *think* so, and yes, the opengraph stuff
- blueoof, vercel
- dreamrealI've got two pri 1 fires to put out right now, so will deploy ASAP but not immediately
- bluebuddy, you got yourself in some deep manure there
- blue(vercel makes nextjs :P)
- dreamrealAll the OG stuff was thrown together. Maybe the fix for teh issue is "rip out the OG content and do it better" but I need to provide better directions for it - JS image manipulation is not something I know well
- bluewhat is this specifically used for?
- dreamrealat delos they assigned images to a junior engineer, who managed to get image generation for frames down to 120ms/per, at 16 frames per set
- dreamrealOG? Link previews on sites
- bluethe og: tags?
- dreamrealI got angry at them and yanked it away from the junior (who didn't do a bad job) and got all 16 frames down to 8ms
- bluelol
- dreamrealbut I did it with SVG because I'm not an idiot
- dreamrealand yes, the og: tags
- bluemein gott
- dreamreal(the junior did not know about SVG, nor did the CTO who assigned it to him: he was like "that won't work!" and insisted on that even after I showed him, until the guy who hired ME told him to shut up, he was embarassing himself)
- dreamrealso... I didn't actually gain any *actual image manipulation* experience because we didn't need images and apparently nobody else realized it
- blueya
- blueconst unbundled = (name: string) => import(name);
- blueew ew ew
- * dreamreal sighs. I liked that CTO but he did not understand programming or why you'd reason about code: he designed a language, but was utterly committed to it even though it had no BNF and no BNF could be constructed for it; he also didn't know about standard programming operators, so he'd CHOOSE operators if they were available, which is why BNF was ... difficult
- blueseriously, who wrote src/og/cards.ts?
- dreamrealblue: I think that'd why node_modules is necessary there
- bluethis majorly skirts primate's bundling
- dreamrealin ui-primate?
- blueyes
- * dreamreal points to claude
- * blue hits claude
- bluethat is TERRIBLE code
- dreamrealHey, I've been saying I threw it together!
- dreamrealI AGREE! How fix pls suh
- blue"Cannot fork because streampack-dev/ui-primate has forking disabled"
- dreamrealHold on, damn it
- bluethis should be on repopack, gh is so hard to work with
- dreamrealgrrr, it won't even allow me to alter the forking. Maybe it needs to be public.
- blueyou can also give me push permissions, and I'll push to rework-og or smth
- blueI just don't wanna push to master directly, I don't like that
- dreamrealIt's public, you should be able to do branches now
- bluek
- dreamrealand I don't want modules to live on rp AND GH, it needs to be a full jump for the entire project
- bluek
- bluekinda pity though you must make the project public just for someone to fork
- blueweird restriction
- dreamrealGH sucks
- dreamrealGL sucks
- dreamrealbut they're better than alternatives like "nothing at all"
- bluewell, RP is coming
- blue"Full suite: Fails during setup because no container runtime is available for the PostgreSQL Testcontainer."
- blueshould I fix your tests whilst at it?
- bluethey currently fail
- dreamrealoh, primate needs the testcontainer, doesn't it
- blueI guess because they assume stuff
- dreamrealI have postgres running locally at all times, and the ui modules were always sort of meant to be private modules, as you can see
- dreamrealso yes
- bluek
- bluedreamreal: are you interested in moving this node impl to rcompat? it currently hardcodes node: stuff, which makes runnable on bun, but not on native paths
- bluethat also means for you that moving ui-primate currently to bun won't speed up anything
- dreamrealI have absolutely NO objection to doing so, and I have no interest in moving to bun
- bluewell, I don't need to put in the work if cross-compat isn't a goal for you
- dreamrealI have no special OBJECTION to it, but eh
- dreamrealwell, here's my thought THERE: I do not mind a single damn bit for this to be a proofing of the projects it uses
- blue@rcompat/* stuff will possibly - but not guaranteed - be accessible to Dog native paths in the future
- dreamreallike, you see how I file bugs as I come across them, to pudl and to primate
- dreamrealand all of that's because the problem isn't originator's-problem-shaped, which is how to find the edges and limits
- blueI think Dog won't rewrite npm stuff unless it's evoked with --rewrite-rcompat or so
- dreamrealall I care about is that the project *works*
- bluewell, you know I don't like putting in work for stuff that doesn't matter to the owner
- blueif it doesn't matter to you, it shouldn't matter to me
- dreamrealwell, making it work well DOES matter to me, and giving originating projects data ALSO matters to me
- dreamreallike, fixing the head stuff? That makes primate better, so that's a total win
- blueI'm fairly agnostic on whether you use native node: paths or rcompat
- dreamrealbut I do not ask for free work
- bluemhm
- bluehttps://github.com/streampack-dev/ui-primate/pull/83
- blue"Primate 0.42.2 now handles these dependencies correctly: with ordinary static imports it bundles Satori into build/server.js and copies Resvg’s native addon to build/native/. The old workaround is obsolete, so the separate runtime node_modules installation can also be removed from the image."
- bluevery good, dreamreal
- bluevery good
- bluenext PR coming your way
- blueI told ya primate can copy native addons :P
- blueI specifically engineered it to do that, to avoid node_modules
- bluehowever, that dynamic import was black voodoo magic
- blueI wonder why claude did that
- dreamreal*nod*
- dreamrealProbably learned from millions of idiot JS programmers, all yapping at their keyboards stupidly
- dreamrealand I don't know JS well enough, being quite honest, to review the code and say "no, don't do that"
- blueya
- dreamrealat delos I LIVED in JS but that was all IOT and device interaction and a completely different profile of problems
- blueya
- bluewell PR is updated, and as I can see, now merged
- dreamrealI'm testing it locally before pushing live
- blueya
- blueI also found out why claude did it
- dreamrealjust run fails with five build errors: `client/annotate.ts:11:25: ERROR: Could not resolve "@streampack-dev/clientjs/annotate"`
- dreamrealwhy did claude do it?
- blueit claims primate can't bundle satori
- dreamrealhaha
- dreamrealwell, that would explain it, yes
- bluedo `npm ci`, then `just run`
- dreamrealgrrrr, damn it to hell
- dreamrealthank you
- blueyw
- bluejust to be fair to claude, it did find an issue with some wasm stuff back in satori 0.32. but that was 2 months ago, your primate is newer. it might have relied on old satori package data
- blue("it" meaning my codex, not claude)
- blueanyway, let's see if it works
- blueI at least can't see that it has issues bundling satori locally
- dreamrealDoing a full release and deploy, can revert if necessary
- bluek
- dreamrealYou reverted the dockerfile, yeah?
- dreamrealor fixed it, rather
- blueyes, I simplified the containerfile back to node_modules-free
- dreamrealgrazi
- dreamrealtodah
- dreamrealsorry
- blueyw
- dreamrealpushing image now
- dreamrealyep, works like a champ
- dreamrealblue++
- nevetblue now has karma of 4.
- bluegreat
- dreamrealis live :D
- bluefyi, that makes your build *portable*. you could use one server for building primate, and another for deploying it. you probably use the same server, but primate allows for a split build/deploy flow
- bluethe important aspect is that the deployment server/image would no longer require npm
- bluecool
- dreamrealI do not build on the server it runs on
- dreamrealand ideally we'd have a multistage build anyway
- bluewell, better, now you don't need to copy node_modules across the world
- dreamrealYou were around when I threw the primate build together, and it was VERY literally thrown together
- blueya
- bluedreamreal: release artifacts are COMING.
- dreamrealnice. Working on a pudl refresh.:/
- bluewhy :/ ?
- blueI thought you liked pudl
- dreamrealI do, I'm refreshing it to meet the pudl design and releases
- dreamrealThe pudl design was based originally on pudl 0.41 and now it's at 0.69.3
- dreamrealI have a lot of workarounds for features pudl actually supports better now
- bluewow
- * blue still doesn't know what pudl is
- blueexcept it can be friends with dog, of course
- bluedreamreal: I need a dedicated server for dog :/
- bluethe vs-v8 benchmarks on my comp keep getting tainted on account of other work
- bluemaybe I'll try to use my other comp
- dreamrealpudl is two things: a UI spec and an implementation of it
- dreamrealit's designed around accessibility for specific types of apps
- dreamrealIMO it's not so great for pure presentation
- blueand the implementation, it's pure frontend?
- dreamrealyes
- blueso why not add a @primate/pudl ?
- * dreamreal shrugs. No reason not to, IMO
- bluewhat's your backend on the pudl client?
- dreamrealfor BCN the primate front end IS tuned for production, where pudl is tuned for an application view
- dreamrealkotlin+spring
- blueyou talk directly to the api, or is that ANOTHER spring app in between?
- dreamrealthe ui-pudl is a spring app itself, a go-between with the UI
- bluemein gott
- dreamrealso postgres <-> app <-> ui-pudl (kotlin) <-> pudl on the client
- bluedouble'd spring, that's a memory nightmare
- dreamrealthe UI is not going to talk to the api directly
- dreamrealnot really
- bluewhy not?
- dreamrealmemory usage on that container's pretty low
- dreamrealwhy would it be a nightmare?
- blueI got (an older version of) nevet on the repopack vps, it's like 685M
- bluememory hog to no end
- dreamrealnevet's the backend though
- blueyes
- bluethat's the 'app' in your <-> chart, yes?
- dreamrealyes
- blueso what has a memory usage that's "pretty low"?
- blueui-pudl (kotlin)?
- dreamrealui-pudl
- bluehow much?
- dreamreali have it set pretty aggressively, it's around 300m, the primate version's way lower, but 300m on that server is well within constraints and could be dropped
- dreamreala lot of that is teh base image itself
- blueomg
- dreamrealme shrugs
- blueyou know repopack is 269mb, right?
- blueand it's not even optimised!
- dreamrealsure
- blueanyway, I think should be possible to create a @primate/pudl, though the gains mightn't be beyond less memory use
- dreamrealsure
- blueunless it has like ssr and friends
- dreamrealI want to say the author does all his backend stuff in go or C#
- bluewhich author
- bluealso, why does pudl release a very like every few hours
- bluethat's such churn
- blueI don't think I'd be able to keep up with that
- bluea version*
- dreamrealclaude
- dreamrealof course
- dreamrealhe works for a POS vendor, I think
- dreamreallike, point of sale
- blueyes
- dreamrealUsed to work for NCR, and Samsung
- bluesounds like a korean
- bluedo you know the guy?
- dreamrealFloridian like me. I've known him since I was 14. :D
- blueoh
- dreamrealhe was the best man at my wedding.
- blueoh WOW
- blueoh goddamn it
- blueI KNOW this guy, he lives in singapore
- dreamrealyes
- blueyou spoke about him before, I'm 100% certain
- dreamrealyes
- dreamrealI am too
- blueok, now it's falling into place
- dreamrealfor him, pudl's research+application, which is how I'm using it, too :D
- bluewell, as we all know, the best people are from florida
- * blue looks at perigrin
- blue- Full clean: 0.26 s vs 0.43 s — about 40% less wall time
- blue- Repeat install: 0.06 s vs 0.11 s — about 45% less
- blue- Production closure: 0.06 s vs 0.11 s — about 45% less
- bluedreamreal: flog vs pnpm, on the primate repo
- dreamrealhah
- dreamrealnice
- dreamrealclosing in on the pudl refresh
- bluenice
- dreamrealvery much a change in UI design
- bluealso: rcompat full API docs are coming
- dreamrealVERY nice
- dreamrealI'm using the pudl refresh to avoid fixing some severe requirement mismatches at $work
- bluecool
- dreamrealI actually told them today that i was rejecting the request with "the speed of light disagrees with this requirement"
- dreamreal"can we make our 8-cpu system emulate a cray with 800 cpus?"
- dreamrealum, no. You need 800 cpus to do that.
- blueoO
- bluelol
- dreamrealWe built the module on a specific set of requirements, and overshot them by 50%... and from the user: the actual requirements look NOTHING like our stated expectations, by like 15000%
- dreamrealyes, I typed the right number of zeros
- blue... and you need 800 cpus
- dreamreal"and oh can we make it faster?"
- dreamrealLike, "we built this to handle 150k inputs, and the requirements said 75k was likely" "cool, here's a 4T input and BTW the search vector isn't 800 but more like... uh... 24000"
- dreamreal"and oh yeah, you take 3ms for the 75k, can we do that 4T input in ... we'll have some grace, 5ms?"
- dreamrealme: "no."
- dreamrealnote: we have CSRs TALKING TO THE CUSTOMERS THE WHOLE TIME.
- dreamrealHow in the effing hell does this kind of scale thing not get brought up
- dreamrealYou might describe me as vaguely unhappy :D
- bluelol
- bluethis is the reason it's hard to trust clankers: const workspace = fs.ref(new URL("../../../", import.meta.url).href)
- dreamrealheh
- bluethey produce a lot of this nonsense. rcompat has a *funtion* for htis, in another package
- blue@rcompat/runtime.projectRoot(import.meta.dirname);
- blueis the correct abstraction
- bluethough import.meta.url should equally work there
- bluebut hardcoding paths.. that's bad
- dreamrealaye
- bluedreamreal: https://repopack.com/project/rcompat/tasks/64
- bluethis is legit cool..
- bluefs.spool... gets you a stream, but ONLY if it passes a digest.
- blueidea: to check the digest, you need to read the stream anyway; this operation saves you the trouble, when you know the expected digest up ahead
- blueyou can tell rcompat, I want this stream, but also, it must satisfy this digest, otherwise error out
- bluethat's an operation that's completely missing node
- bluefrom node*
- bluehi phaleth
- bluehaha phaleth
- blueyou lashing out at clankers is the funniest ever
- blue" \Stop blocking the logo and the favicon because of CPS already and go do something useful."
- blueLOL
- phalethhi blue
- phalethyeah
- phalethwhy did clanker add CPS to the docs anyway
- blueno idea
- blueit's flog
- bluebtw, I merged this earlier https://repopack.com/project/flog/tasks/80
- bluebtw, https://repopack.com/project/rcompat/tasks/61
- bluebtw, https://repopack.com/project/rcompat/tasks/62
- bluebtw, https://ptx6lcpc3-task-65.repopack.app/api/fs
- nevet@rcompat/fs API · rcompat
- bluedreamreal: ^
- bluesystemfault: ^
- blueincluding hover shenanigans!
- phalethlets try that linux speedup
- bluethese API docs are amaze sauce
- bluethey also unveiled some private types that should be public
- bluesome music for chill https://www.youtube.com/watch?v=KLlXJH7oJSI
- nevetYouTube: Tocotronic – Freiburg V3.0 (Official Video) | Tocotronic
- phalethflog docs site is peak performance, rcompat docs is an app pulling in a lot of JS to the browser
- blueawww
- blueit also takes a truckload of time to redeploy, gets on my poor nerves
- bluehttps://ptx6lcpc3-task-65.repopack.app/api/symbol
- nevet@rcompat/symbol API · rcompat
- phalethbut why is that on site there is no app and on another kinda same there is the app JS bundle?
- bluewhere?
- blueyou mean in flog
- blue?
- blueOMG THAT LOGO
- bluephaleth you are a genius
- phalethyeah, take a look at https://flog.run vs. https://rcompat.repopack.app
- nevethttps://flog.run "flog — predictable JavaScript package management" || https://rcompat.repopack.app "rcompat — JavaScript standard library"
- phalethheh
- bluethe rcompat website is seriously messed up
- bluewth
- bluealso I'm not getting a favicon on flog website
- bluealso, we probably need to adapt the flog website colours to the logo
- bluedamn, this logo is too good
- phalethand on rcompat docs you get the favicon?
- blueyes
- bluebtecause I wired the damn thing myself
- phaleththen LShift + Refresh
- blueyeah, this is a private brave tab
- phalethnah
- bluei'll close it
- phalethlet me check
- bluegoddamn brave
- bluenah, not getting favicon'd
- phalethso you put in <link rel="icon" href="/logo.svg" type="image/svg+xml" />
- phalethand then it works on brave
- blueyes, like a big idiot
- blueand primate website works without it
- blueI have no idea why
- blueI hate browsers
- bluebrowsers suck
- phalethI can see the favicon no problem
- * dreamreal hands blue w3m, lynx, links, gopher
- bluemaybe it's just brave, idk
- phalethand I've never seen <link rel="icon" href="/logo.svg" type="image/svg+xml" /> on any website
- bluedreamreal, can you see favicon https://flog.run/
- nevetflog — predictable JavaScript package management
- blueoh yeah, tell me about it phaleth. dunno, clanker suggested that
- dreamrealno
- blueWHY AREN'T THE FAVICONS SHOWING UP
- dreamrealhas the brave browser's default "world" favioc
- bluefriggin brave
- blueI'm gonna try another browser
- blueI gotta know
- blueI might have a chromium flying around here somewhere
- dreamrealsafari... looks like it's rendering SOMETHING?
- blueHAH
- dreamrealit's too small to tell what it is
- blueIT WORKS ON CHROMIUM
- blueI KNEW IT
- * blue kicks brave
- dreamrealchrome works too
- dreamrealsafari != chromium, though, I thought. Webkit, right?
- phalethI bet safari does not show the favicon
- blueya. safari is webkit, chrome is like the google thing, whatever it's called
- phalethcause epiphany doesn't
- bluesafari webkit/jsc
- phalethchrome is v8 and blink
- bluechrome is whatever the h the google renderer is called and v8
- blueyes, blink, thank you
- blueand why can't brave show the favicon now?
- blueonly god knows
- bluewe're only in 2026 and we haven't solved yet favicons
- blueI'm gonn adrop this explicit <link thing in rcompat
- bluethis is ridiculous
- phalethcan brave show favicon from bytecode.news?
- phalethI can add some metatags, but I'd like to avoid that
- blueYES
- blueit can
- bluea b over a black background with a red corner
- bluebut damn it dreamreal, that looks like...
- blueyou need phaleth for a new favion
- phalethok, I will try to fix that later for brave then
- bluehe's an aritst
- bluephaleth where is the favicon for dog
- phalethdog doesn't have a website
- bluewe all know favicons are the most important for a project
- phalethdog doesn't even exists I hope
- phalethyeah
- blueya but dog exists
- bluehttps://repopack.com/project/dog
- blueI hope dreamreal doesn't die of laughter here
- phalethso it's like a secret project
- bluewell it's secret until it reaches parity with v8, yes
- bluecurrently it's 9x slower
- blueit started at 30x, so we're going somewhere
- bluethe real goal is 5x faster than v8, though
- phalethlet me give you something to read
- blueI read a whitepaper over go jit perf earlier
- bluethat went nowhere, unfortunately
- blueanyway, if you fix the flog favicon on brave, fix that rcompat dumpster fire too
- bluefavicons should be auto discoverable via favicon.ico, not in <link
- phalethhttps://encore.dev/blog/rust-runtime
- nevetWhat We Learned Building a Rust Runtime for TypeScript
- phalethyeah, sure
- bluephaleth: can you have js.stdlib.run point to rcompat.repopack.app, please?
- bluejust registered
- phalethyeah, sure
- bluethx
- bluejs.stdlib.run 207.180.197.249 600
- phalethfavicon shows up twice on palemoon https://i.imageupload.app/7c3a342918e297635996.png
- bluerecord is there
- phalethalso the site is fast on palemoon cause it's peek performance
- bluedog will be hosted at dog.flog.run, fyi
- phalethhope you didn't pay $8k for that domain
- blue$4
- phalethok, nice
- phalethanyway, why js as subdomain?
- bluebecause it's a js stblib
- bluestdlib*
- phalethok, well, then there will be a TLS cert jut for that, but why not use stdlib.run also?
- phalethat least temporarily
- phalethand www.stdlib.run for that matter as well
- bluewe can use stblib.run if you prefer
- blueI'm not dogmatic on that, I just thought js. is cooler
- bluedreamreal: opinion?
- bluejs.stdlib.run or stdlib.run ?
- phalethok, I will setup all three of them with the same TLS cert
- dreamrealI'm all for it! What is the opinion about?
- blue20:15 < blue> js.stdlib.run or stdlib.run
- dreamrealoh, js.stdlib.run
- dreamrealit's scoped to js
- dreamreal... said with no forethought whatsoever on my part
- bluephaleth: we can redirect stdlib.run and www stblib.run to js.stdlib.run
- phalethok, I will setup the rewrite
- blueA js.stdlib.run 207.180.197.249 600
- blueA stdlib.run 207.180.197.249 600
- blueCNAME www.stdlib.run stdlib.run. 600
- bluegood?
- blueI thought of CNAMING directly to js.stdlib.run, but nah, sounds weird
- blueanyway, all three should point at rcompat.repopack.app, methinks
- blueor rather, the non js. ones should probably redirect to .js
- bluejs.*
- blueyeah, probably best
- phalethyeah
- bluethen dreamreal could send his buddy a real nice domain
- blueand his buddy won't have reservations anymore, clearly
- bluebecause domains are all that matters
- dreamrealwell, the ability to understand is relatively important
- blueya
- blueI should be able to finish the API docs today or tomorrow, and merge htem onto master
- blueand together with the new domain, it's gonna be wicked
- bluethen I'm gonna fix the flog website and throw out all the nonsense clanker garbage about primate
- blueout-of-the-box thinking is not a particular strength of clankers, I find
- blueyes, I used primate as a reference point for flog working as intended, that doesn't mean you need to write A WHOLE SECTION about it into the docs
- blueidiot
- dreamrealyes, they take offhand suggestions and explanations VERY seriously
- dreamrealI think I wrote an article about that
- blueI think I read it
- dreamreali think so, too. :D
- bluephaleth: btw, I fixed the node_modules pulling-in on BCN's primate client, earlier. clanker hallucinations
- bluenow it's pristine, and dreamreal's server is spared from sinking into the black hole known as node_modules
- dreamrealI find "hallucination" to be the wrong term - they're not hallucinations as much as model inferences
- bluewell, in that it decided that primate can't bundle .node addons, which is specifically NOT the case
- phalethblue: oh, I was gonna look into that, so you fixed that in primate?
- dreamrealit inferred, because it has no judgment
- blueno, I fixed it in BCN/primate
- dreamrealhe fixed it in the build
- blueas in, ui-primate, I mean
- phalethok, cool, I will see what you did
- dreamrealit was my code, not primate's, because *I* didn't write the primate UI, I directed it
- blueclaude really outdid itself there
- bluepeak stuff
- phalethor what clanker did anyway, curious
- blueit tried (and managed) to bust primate's bundler by inventing a dynamically importing function
- bluebecause it was of the opinion that primate can't do something
- blueproblem if, there are very few things, and still fewer USABLE things, that primate's bundler, esbuild, CAN'T do
- blueproblem is*
- bluehttps://github.com/streampack-dev/ui-primate/commit/5697be1ad785ba423b8a9ead3738756fedcc9615#diff-7bd1e06f12808abf8ebcab5b3789bc58f25461cc27e79f616ea5fceb84b48af6L9-L11
- nevetMerge pull request #83 from terrablue/master · streampack-dev/ui-primate@5697be1
- blue"satori's WebAssembly and resvg's native binary don't survive Primate's server bundle"
- bluethis is a lie
- bluehence hallucination
- blue"and Primate has no way to leave a package out of it"
- bluethis could have been true at the time. I believe it arrived in 0.42, and ui-primate was written against 0.41
- bluehowever, irrelevant, because the capability wasn't necessary
- bluehence this amazing code: const unbundled = (name: string) => import(name);
- bluewhich esbuild can't unroll
- blue>_>
- bluehttps://ptx6lcpc3-task-65.repopack.app/api/type
- nevet@rcompat/type API · rcompat
- blueis now ported
- blueI gotta do something about these terrible task urls
- perigrinfloridia does represent ... as do ex samsung employees
- bluelol
- perigrin*Florida (see you can tell I'm a product of Florida's education system)
- blueYou're a Florida Man :P
- perigrinI have my moments.
- bluelol
- bluephaleth: I love the flog logo
- bluewhen I'm done with the rcompat website I'll prop up the flog website
- phalethI had to put temporary redirect there cause otherwise browser will not change the hostname of the url https://stdlib.run/
- nevetrcompat — JavaScript standard library
- phalethheh, ok, I had fun making that logo
- phalethactually I think the colors could be better, but whatever
- phalethblue: anyway, try to figure out why is the rcompat docs site a full blown app, that's kinda weird
- blueya will do
- bluephaleth: peak rcompat website coming
- phalethnice
- blueThe workaround now produces fully static pages:
- blue- /api/array: 16,569 → 6,873 bytes (−59%)
- blue- /api/assert: 186,290 → 75,266 bytes (−60%)
- blue- Browser JS: 85.8 KB → 0
- blue- Hydration/walker scripts: 3 → 0
- blue- Page responses now include Cache-Control: public, max-age=300, must-revalidate
- bluephaleth: good? ^
- bluealso seems to have found a Marko <Head> bug, but perhaps dreamreal has already reported, unsure if this is the same thing
- phalethawesome
- blueI reported upstream to the primate team, hope they have some competent people to fix it
- bluehttps://repopack.com/project/primate/tasks/232
- bluethis isn't relevant I think to dreamreal, he uses hydration
- phalethyeah, bytecode.news uses the app
- blueanyway I don't think the primate team will backport it to .42
- bluenor doe the rcompat consider it necessary
- blueso it's all good
- blueprimate#232 is being worked on
- phalethmarko({ csr: false }) does the trick, nice
- blueyes
- bluebut still, the <Head> tag should play nicely without hydration
- blueand not force a workaround
- bluethe primate team is on it, but it won't get out until .43
- phalethcan you do a deployment preview on https://repopack.com/project/rcompat/tasks/65 ?
- bluehttps://ptx6lcpc3-task-65.repopack.app/
- blueit's building right now
- phalethoh, it was already there, it says No Preview on the task page for me
- phalethgreat, that's super fast
- blueya, I might need to fix permissions there
- bluethere you go
- bluefixed
- phalethyeah, now I can see the preview link
- blueyeah, very fast
- blueignore that terrible link. it'll look better in repopack ui once I have a bit of time
- blueshould be a button, not a link
- blueoh man, those websites are FAST
- phalethit's an anchor tag
- phalethbut that's good
- blueya
- bluenice, js.stdlib.run is online!
- blueonce I'm done with the api stuff I'll merge it and then it's all fine and dandy
- phalethso it's server rendered from TypeScript?
- blueyes
- blueit's hard to get dreamreal excited about stuff: so if I managed to do it, I know I'm on the right track
- phalethshould be fast the, not sure about node
- blueya
- phaleththen*
- bluewith dog will be faster
- bluealso rcompat is still being bundled with pnpm
- bluewhich is going to be addressed soon
- blueanyway, I'm off to play a board game and then chill for the rest of the evening
- blueyou fine gentleman, have a great rest of day
- bluegentlemen*
- bluealso note the wallop by libera team, and update your browsers/systems if applicable
- bluesomething something certificate keys leaked
- phalethok, bye
- dreamrealblue: let's talk in the morning about the zig article
- bluedreamreal: ok mate
- dreamrealAre you happy with where you left it?
- bluemostly, yes. I've kept some distance for a final human pass, it's still a bit too clankery
- bluefor my taste
- bluemostly thinking about std loaders right now, dreamreal
- bluewhether I want to do @rcompat/{yaml,toml,json5,jsonc}, or skip for now. bun offers most of them native, and it's something that I think would be good to have in a stdlib
- dreamreal*nod* I think the next question for that zig thing is what you want to post it as: do you want ME to post it an attribute it to a pseudonym, do you want me to edit it for voice to remove some of the clanker-y bits, as you put it...
- dreamrealdo you want me to narrate it, etc
- bluedreamreal: I think we can post it under my `blue` username. But I'd be grateful if you could edit it for voice and remove the clanker-y bits, because one of the thing that annoyed me about the last article I wrote on my blog, was how many people complained it was AI-generated. I do NOT want the HN plebs to complain about that - they should engage with content, not tone, but it's very hard for them to do
- bluethat if they smell clankerness
- blueone of the things*
- dreamrealblue: okay, will look at it tomorrow morning
- bluethx
- bluedreamreal: https://js.stdlib.run/
- nevetrcompat — JavaScript standard library
- bluehttps://js.stdlib.run/api
- bluefull API
- bluethere are still a few smaller things to fix, but it's looking good. tomorrow I'm going to publish new versions of everything
