The Gram Code Editor - 2026-03-27


  1. nicoco

    the investors are happy to see new releases anyway, kpis go brrrrrrrr

  2. krig

    šŸ˜„

  3. nicoco

    what is the icon theme used on the website frontpage?

  4. nicoco

    https://chat.slidge.im:5281/file_share/069c6484-d9b0-7214-b51f-05e11c7529ce/06dd749f-3f82-453c-a46e-c2c46a43380d.png

  5. nicoco

    ^ this one

  6. krig

    I think that’s one of the Catppuccin ones? Not sure, sorry!

  7. nicoco

    since icons are off by default, maybe that screenshot is a bit misleading

  8. nicoco

    why that choice btw?

  9. krig

    More of a mistake than a choice. I spent like six months just messing around with the code, and apparently changed the defaults at some point. But in my own config, I had turned on icons again so I didn’t notice

  10. nicoco

    very much a detail, but I believe the screenshot on the front page should reflect the default?

  11. krig

    and now I am reluctant to change the defaults once people are using the editor. But I just pushed a change for the next version which will turn on icons if they are off by default and someone changes the icon theme

  12. nicoco

    ok!

  13. krig

    Maybe. Not sure. The configurability is one of the features I want to highlight..

  14. nicoco

    fair enough. those catpuccin icons look nice, I just switched :)

  15. nicoco

    btw I had a look at RPM package registry in codeberg, it looks very similar to debian, and since my work-provided laptop is running Fedora I think I'll submit a small PR for this so my usual `dnf update` updates gram too :)

  16. krig

    I think what I’ll do when I revise the website is to make it a carousel that shows off different features. There is a PR for that already! It just needs a typo fix to merge

  17. krig

    by theDoctor

  18. krig

    for RPM registry that is

  19. nicoco

    ha ok well, less work for me then :)

  20. nicoco

    ha damn, the new rendering thing is causing issues on sway/intel (very annoying flickering randomly occuring). previously I worked around that by using `env -u WAYLAND_DISPLAY` which I think forced x11 through Xwayland, but 1.2.1 just hangs if I do that :(

  21. nicoco

    logs stop as INFO [gram] ========== starting gram version 1.2.1, sha 0b406ac ========== and the process is `[defunt]` according to ps

  22. krig

    :/ I should have held off on the wgpu merge.. especially since I don’t use Linux that much. Maybe I should release 1.3.0 and revert it, and make it a branch for now. Sigh..

  23. nicoco

    I think I've read something similar in an issue thread, but I see that gram uses (?) my nivida card: [gram::gram] Using GPU: GpuSpecs { is_software_emulated: false, device_name: "NVIDIA RTX 2000 Ada Generation Laptop GPU", driver_name: "NVIDIA", driver_info: "580.126.18" }

  24. krig

    But did that env trick work in 1.2.0? Because there are no code changes at all between 1.2.0 and 1.2.1

  25. krig

    ah, right.. you can force it to use a different GPU with GRAM_DEVICE_ID

  26. nicoco

    oh let me try that GRAM_DEVICE_ID

  27. nicoco

    I did not even know it was possible to use the nvidia gpu for display using sway xD

  28. nicoco

    (I only use it for CUDA stuff, yes I do machine learning, please don't ban me)

  29. krig

    I guess it isn’t šŸ˜…

  30. krig

    no evil stuff I hope

  31. nicoco

    > But did that env trick work in 1.2.0? Because there are no code changes at all between 1.2.0 and 1.2.1 mm no I was using < 1.2.0 on this laptop previously

  32. nicoco

    > I guess it isn’t šŸ˜… well yes it is, but it's flickery xD

  33. nicoco

    no evil stuff, I hope so, machine learning for medical image segmentation. not the fancy newest architectures like transformers, boring "old" convolutional networks :)

  34. nicoco

    How do I know which GRAM_DEVICE_ID value I should set?

  35. krig

    I used to study AI before LLMs, when it was more alpha-beta pruning and less data centers on fire. Had a chess app in the iOS store early on.. I’m mostly frustrated that the whole field got taken over by finance scammer bros.

  36. krig

    ā€lspci -nn | grep VGAā€ should tell you the device id, something like ā€[10de:2484]ā€. Then set GRAM_DEVICE_ID=0x2484

  37. nicoco

    Well, I think we should not throw away the baby with the bath water, machine learning has non-evil use-cases for sure. The hype is crazy and destructive, but eg, for this very specific task I mentioned, segmentation of medical images, there is no match, all other approaches are inferior. And with segmented medical images you can do a lot of fun (and maybe useful…) things. Who knows, maybe I'm wrong :)

  38. nicoco

    > ā€lspci -nn | grep VGAā€ should tell you the device id, something like ā€[10de:2484]ā€. Then set GRAM_DEVICE_ID=0x2484 that worked to force intel (interestingly, the nvidia card was not listed by that command). I'll see if the random flickering is still there after a few minutes of usage I guess :)

  39. nicoco

    thanks!

  40. krig

    No I am sure there are legitimate uses. But right now those are getting crushed by the illegitimate ones. The first practical application I saw of machine learning in the wild was a company which was building face recognition into vending machines, so that they could detect gender and race of the person shopping. Just completely amoral, horrible shit.. I just don’t think our society is equipped to handle this technology responsibly.

  41. nicoco

    (damn it flickers too0

  42. nicoco

    > No I am sure there are legitimate uses. But right now those are getting crushed by the illegitimate ones. The first practical application I saw of machine learning in the wild was a company which was building face recognition into vending machines, so that they could detect gender and race of the person shopping. Just completely amoral, horrible shit.. I just don’t think our society is equipped to handle this technology responsibly. no need to convince me about that

  43. krig

    I had that problem or something similar on my ARM SoC before the wgpu move. It’s probably bugs in the editor and not the graphics library then..

  44. nicoco

    To play devil's advocate, I think a lot of radical criticism of "AI" (gosh I hate that marketing terminology) applies to computers (or even technology) in general. And maybe it is valid and computers are evil and I should burn all my electronics and grow tomatoes instead. I have not made that jump yet :)

  45. krig

    Totally. Some days I just want to farm potatoes and never look at a screen again, haha

  46. nicoco

    well, gram does not help for that, since I enjoy using it a lot =)

  47. krig

    oh no šŸ˜„

  48. nicoco

    anyway I reverted to 1ab182fb7f78ff86a77c0585cad2564a96267224 which I had previously built and allows forcing Xwayland which does not flicker

  49. botaki

    > :/ I should have held off on the wgpu merge.. especially since I don’t use Linux that much. Maybe I should release 1.3.0 and revert it, and make it a branch for now. Sigh.. Yes maybe wgpu should be reverted and added when more debugging and testing is done.

  50. krig

    so there’s a bug in sway that may be the issue here actually: https://github.com/swaywm/sway/issues/8755

  51. nicoco

    oh

  52. nicoco

    I'll live with an old gram version until this is fixed and released (or maybe I'll try these env var workarounds when I'm bored ^^)

  53. krig

    I can create a branch that is 1.2.1 with the wgpu patches reverted, shouldn’t be too difficult to put together

  54. krig

    and build a release from that

  55. nicoco

    whatever you feel is best. I am not a real software engineer but one thing I've learned about it over the years is that you can't have workarounds everywhere to ensure compatibility with buggy third parties

  56. nicoco

    if it's a sway bug, it's a sway bug and this is where efforts should go (however, this is completely out of my depth, I can't help xD)

  57. krig

    no, it wouldn’t be a long term thing.. I still have hope to get the wgpu stuff in better shape eventually

  58. krig

    I find it kind of ridiculous to have these kinds of issues with a text editor to be honest :D even if it wants hardware acceleration, it should run just fine on some ancient opengl

  59. nicoco

    Well, I trust your call on that.

  60. krig

    but yeah that’s the editor I forked, so that’s what we have to work with..

  61. nicoco

    I mean, perf is good, but coming from now enshittifAIed jetbrains IDE I can't imagine rendering text on the screen would ever be a performance bottleneck I could feel

  62. nicoco

    (but I don't know much about the subject tbh)

  63. krig

    it’s silly yeah

  64. thedoctor

    > ha ok well, less work for me then :) I hadn't looked in here a while and it occurred to me that it wasa logical thing to do šŸ˜… Not sure what to think of the signing issue, though.

  65. thedoctor

    Speaking of packages: there are now three AUR packages for gram and there's one that I think is better quality to use for us "gram-editor-bin". It does list a few things as runtime dependencies that I don't know are necessary, though. Things like openssl, sqlite, alsa-libs. Any thoughts on that?

  66. thedoctor

    Speaking of packages: there are now three AUR packages for gram and there's one that I think is better quality to use for us, namely "gram-editor-bin". It does list a few things as runtime dependencies that I don't know are necessary, though. Things like openssl, sqlite, alsa-libs. Any thoughts on that?

  67. krig

    I think @nerdyslacker who created that one is working on another version

  68. krig

    so for the signing, was that a missing feature in forgejo or something we need to do? hmm

  69. krig

    ah I see, codeberg doesn’t enable it. Guess we just have to wait for them to turn it on

  70. thedoctor

    > I think @nerdyslacker who created that one is working on another version I tried adapting one of the PKGBUILD files, built a package and uploaded it to my fork of gram, pretty straightforward in total. I did it inam Arch container, though. Is it possible to have one in CI with Codeberg somehow?

  71. thedoctor

    > I think @nerdyslacker who created that one is working on another version I tried adapting one of the PKGBUILD files, built a package and uploaded it to my fork of gram, pretty straightforward in total. I did it in an Arch container, though. Is it possible to have one in CI with Codeberg somehow?

  72. krig

    yeah that should be possible, I need to add another base image to the runner

  73. krig

    I can look at that yeah

  74. thedoctor

    Just submitted a PR adding build instructions for an Arch package. Once this is built in CI, I would add the necessary steps to push the package into the Gram package registry and add instructions to add it as a custom Arch repo.

  75. krig

    Hi, yes I have seen it, will check it out as soon as I can!

  76. thedoctor

    > Hi, yes I have seen it, will check it out as soon as I can! No need to rush, enjoy your weekend :)