Community
Chat Logs
Wednesday, October 14, 2026 (America/New_York)
- dreamrealhttps://bytecode.news/posts/2026/10/where-is-all-the-new-good-software
- bluedreamreal: maybe Nate Meyvis should be looking at primate.run, flog.run or repopack.com
- dreamrealnot his point
- dreamrealthose are drops of water in a massive ocean
- bluehis point is flawed. no one is making the claim that "AI skeptics" are looking to refute
- blueyou can't prove a negative, the guy fails 101
- bluelistening
- * dreamreal sighs
- bluealso, he doesn't list omarchy
- bluewhich is widely popular and almost 100% clankered
- dreamrealmein Gott
- dreamrealnone of those matter for what he's talking about
- dreamrealthe existence of 100 new fascinating things is being drowned out by other noise
- dreamrealit's a valid thing to wonder about, even if those 100 things exist
- blueI found a regression in the podcast version, dreamreal
- dreamrealfire away
- blueI can't listen as fast as I read
- dreamrealI changed something in how it was record it, Oh, cool! So don't listen
- dreamrealI can't listen to them myself
- bluewell I prefer listening
- bluethis is a good article
- dreamrealhold on got something to do here
- bluedo you consider repopack falling as a 'reimplementation', or something new?
- bluefalling under*
- bluebecause at first it was mostly a reimplementation of gh+herdr, but now it's really going its separate ways conceptually, with tasks
- dreamrealNo, but why?
- bluejust interested
- dreamrealI mean, I find new thiings all the time
- dreamrealbut then again I *look*
- blueya
- dreamrealI find the written article > the audio for a lot of reasons, not least is that only one footnote made it into the audio
- bluewell this is a good article. which I can't say about the original article
- bluedon't like its form much, bulletin lists strike me as lazy
- dreamrealI didn't think the original article was bad (or else I wouldn't have written it up on BCN)
- bluethat depends on your definition of bad
- blueit's pretty bad for what it tries to do
- dreamrealI felt like he had a good observation that needed some context
- blueI am sorely missing examples, again
- blue"We are in fact seeing software improvement, in the form of more and better features in existing products."
- bluewhere?
- dreamrealSo? I mean, what's better, to run the first mile of a marathon or just to not try at all? If your only metric is "did you finish" then they're equivalent, but I would suggest that the first mile is a START
- dreamrealAre you talking about where *I* didn't point out improvements?
- blueno
- dreamrealBecause I surely did not
- dreamrealwell, he has the same problem I do: endorsement
- blueno I'm just saying, that article is not *good*. it could have good aspects, but I don't find it useful. the BCN take gave us EXAMPLES
- dreamrealhis problem is less severe than mine, I think, but calling out one or three makes it seem like those are THE IMPROVEMENTS
- blueand it was prose, which is infinitely better than bulletins
- dreamrealwell, that's why BCN wrote it up :D
- dreamrealI thought it was a useful thoughtline, but it didn't actually explore the space especially well
- bluethe visibility point, was well delivered. but I can't find it in any accessible form in the source material
- dreamrealyeah, well, it's not there
- dreamrealI think he had an idea and developed it about 50% because it's a blog
- bluealso I think the thing that drives the BCN take home, and imho is completely missing, is the implementation-ideation chasm
- bluebecause that's the golden nugget here
- dreamreal... which is why I wrote it. And I really did ask Fable about that :D
- bluethe problem is real, and kind of acute
- dreamrealI would have quoted it and GOSH did I want to, but its answer was way too long to quote
- dreamrealI was gonna use a vocoder in the podcast version for the bot :D
- blueAI cannot imagine us the good/new or new/good software, because it doesn't have access to wishes and desires, which drive this
- dreamrealyeah, it basically does efficient combinatorics
- bluebut our wishes and devices are infinitely masked by all the bad software out there
- blueI *still* don't understand why I need to scan medical bills, upload them to my insurance, get reimbursed, and transfer the money myself
- bluebut I have been accustomed, or forced, into this broken flow
- dreamrealouch, I haven't had to do that for years
- blueso of course I'm not going to think a better one is possible
- dreamrealI thought you guys had better medical support
- bluewell our private insurance is awesome. but ironically involves a lot of boilerplating, while public insurance involves none, and sucks to no end
- bluegotta choose your poison. wait 5 days for an appointment, but do all the invoicing yourself, or wait 5 months
- dreamrealheh, don't say that in public HERE, all the blue tribe people would hate you for pointing out the problem
- blueit would be nicer though if the doctor could just send the invoice directly to my insurance
- blueI do have like 0.5k of out-of-pocket per year, but I wouldn't mind my insurance billing me that
- blueor just waiting till year's end to see if it's worth billing
- bluebut generally speaking here, the private insurance does not interact directly with the doctors
- dreamreallots of BCN changes queued today, mostly QOL stuff, and I'm going to fix startup deployment I HOPE
- blueI pay the doctor, then get reimubrsed
- bluecool
- dreamrealbut there are backend changes, so the UI sync dance gets to happen :/
- dreamrealAll of the visible timestamps are UTC and ew
- bluedreamreal: related, https://repopack.com/project/primate/tasks/211?tasks=status%3Dopen
- bluenote that that is *experimental*. I have yet to ideate how to do that
- bluethe problem is probably known, and yet not easy: you want to be able to deploy migrations *while* your service is running; so you need a bridge situation
- bluein that bridge situation, your code is old, but your db has a newer state
- dreamrealyeah, that's... going to be interesting if the mutation is destructive
- bluewell, a lot of this is related to the fact that git is just not a very good tool for complex stuff, let alone AI stuff
- blueconsider diffs: diffs are most useful *in context*. if a new function is exported from file A, and now used in file B, you want to see this together; but this isn't how git does diffs
- bluethe problem is that git primitives do not lend themselves well at all to those two requirements: mixed-version apps, and context-based diffs
- bluedreamreal: cloudflare has been working on a git replacement for agents. I haven't looked at it, but this isn't the first time I'm hearing git hasn't been serving us well
- bluethis is the same thing as the BCN article noted: we're so accustomed to (bad) stuff, we hardly ever peek outside
- dreamrealyeah, it's hard to think "what's new" as opposed to maybe "how do I make this 1% better" when 1% just isn't good enough
- dreamrealanyway, exercise time
- blueenjoy
- bot[Node.js Blog] Wednesday, October 21, 2026 Security Releases - https://nodejs.org/en/blog/vulnerability/october-2026-security-releases
- bluedreamreal: thinking of ai-forking node
- bluemostly to add support for "std:" namespaces (rcompat)
- blueI'm waiting for the codex refresh, then will embark on that
- systemfaultIs that something that can be "shim"ed or that requires an actual fork?
- bluesystemfault: I don't know... maybe it's enough to change the .js files without touching the c++ part
- bluebut I probably need to touch the package/import resolution
- bot[Node.js Blog] Node.js 24.22.0 (LTS) - https://nodejs.org/en/blog/release/v24.22.0
- bot[Node.js Blog] Node.js 26.12.0 (Current) - https://nodejs.org/en/blog/release/v26.12.0
- bluehi phaleth
- bluedid you hear deno is dead?
- bluehttps://bytes.dev/archives/529
- nevetBytes #529 - A Eulogy for Deno
- phalethhi blue
- phalethyeah, that's why I've looked up node's permissions
- bluethis is done btw https://repopack.com/project/primate/tasks/209
- phalethis this correct?
- phalethRUN flog install
- phalethRUN flog run --cwd apps/website build
- blueyes
- phalethI thought yesterday or two days ago you figured out how to get only the packages needed to build primate website
- blue--cwd is for run, --workspace is for install
- phalethyeah, so why not use --workspace for install is the question
- phalethalso clanker says it installed over 400 packages there, that's too much I think
- blueyou can
- phalethso should I fix that Containerfile?
- phalethit's overcomplicated anyway
- bluechanging
- bluechanged
- phalethclanker should get: installed 361 packages and 65 workspace links (2 cache hits, 359 downloads)
- blueinstalled 361 packages and 65 workspace links (2 cache hits, 359 downloads)
- blueyes?
- blueI think that's still too much, but ok
- phalethwow, flog is so much faster than yesterday
- blueyes?
- phalethfrom 2:45 to 1:15
- blue1:15 minutes?
- phalethI mean the whole image build
- bluestill too long
- phalethyeah
- bluehow long was pnpm/nub?
- phalethnah, my computer is old
- blueah ok
- phalethforever I think
- blueya flog is superfast
- bluephaleth: I'm planning a node fork, called oden
- blueai-driven, of course
- bluethis is a parallel developed to flog. flog will get support for running js via different engines, but will have a different goal
- phaleththat is too much effort
- blueoden will maintain node compat 100%, but also support "std:"
- bluewell we created flog in 4 days, we can do anything
- phalethwhat about repopack?
- bluerepopack is still around 95% of the effort, I've been workin on it all week
- phalethI'd highly recommend leaving node alone, you already have rcompat to deal with shortcomings
- phalethlately it makes steady progress as bun is pushing it around
- phalethalso oden sounds like programming lang
- phaletheh, it just randomly decided to add react types to apps/website/package.json
- phalethand also rxjs
- phalethso I cannot revoke approval
- phalethoh, I can withdraw submission
- blueya
- bluerxjs and the react types were left from an earlier thingy, should be removed
- phalethyup
- blueflog is fastest for you currently, phaleth ?
- bluebtw no cgo
- phalethI think we will see after you merge 209
- phalethI do not have Containerfile with nub or pnpm anymore
- bluek
- phalethbut you could ask clanker in another issue to reconstruct those and run benchmarks and put that on flog site
- phalethcause that's really only thing people are interested in
- phalethmyself included
- phalethanyway, I think it's too much effort, flog is pretty good now
- blueya
- bluephaleth could you create a flog task for those benchmarkos? pnpm vs npm (if applicable) vs bun maybe
- phalethjust don't worry about it, flog is fine
- jreicherblue: do you know about zdiff3 conflict style in git? It's not exactly what you're talking about, but extra context in another sense is what it provides, so I think the problem is somewhat recognised.
- bluejreicher: I don't, pray tell more
- jreicherThe default conflict style has two text sections, one from the destination and one from the source. The zdiff3 conflict style adds a third: the text from the most recent ancestor, so you see what was there before the branches diverged.
- jreichermost recent COMMON ancestor, sorry
- blueI see
- jreicherIn many merge tools this brings up a different view too, and a really, really good one.