>>109552794
Gump. Forrest Gump.

>>109552794
very interesting. can you give me a link to read more about this project?

>>109552794
i got sidetracked and started making a programming language first.

>>109553590
There is none, I don't like writing documentation that much besides the usual function preamble and file headers, but if you're curious I can aswer your questions.

It's a 32bit x86 kernel with some Posix compatibility inside libcrt, but architecturally it borrows more from Windows (especially the window manager, as you can see).
I've been working on it for about 6 years and there are many features to list, I have a small web browser implementation on it, a Mastodon client, a 4chan read-only client with proper filtering, a modern code editor, REST client, IRC client, Mail client (can send and retrieve emails), full TLS support via BearSSL, Usenet client (read and write messages), Tor socks support with proper ntor key negociation but no ".onion" hidden services support for now, socks4/5, etc.

File: Screenshot 2026-07-31 at (...).png (181 KB, 2560x1440)

181 KB PNG

>>109553646
And it's booting on a Sony Vaio P netbook + the usual QEMU for testing, with multiple drivers: basic ACPI, no SMP, USB uhci/ehci/xhci, audio Cc'97 (for testing, Vaio needs HDA), RTL8139 network (for testing, Vaio needs other stuff), the usual PCI, SMbus, I2C.

Let me guess, it's vibecoded.

On the other hand, creating your own OS theoretically has never been easier.
>write an interface for Linux drivers
>write your own suite
>port Chromium to get all the webapps in the world (most non-specialized apps we take for granted)
>port Wine/Proton to get most Windows apps and games
>port some free software and libraries

>>109553917
but where's the fun in that if all you're doing is just reimplementing POSIX/Linux interfaces.
I'd love to see more innovative operating systems, such as TempleOS or Plan9 (albeit it still has a ton of UNIX heritage).

>>109553646
>Posix
its shit

>>109553928
You wouldn't be (other than the drivers) and the fun is that you can actually use it beyond hobbyism.
Like haiku already does most of these things.

>>109553917
It's not vibecoded, took me 6 years to get to this.

>>109552794
>why is it so cool to work on your own OS?
It's not. It's NOT. STOP enjoying things I don't enjoy, sheesh. Didn't you get the memo?

>>109554056
Based

>>109552794
yep, I'm thinking based (or troll)

>>109553928
This. Always heard good things about how VMS is architected and how it differs from POSIX. We need more hobby and alternate OSes that aren't just *nix clones.

vibecoding my os as we speak
its capability based and non-unix and non-posix

File: Screenshot 2026-08-01 at (...).png (165 KB, 2560x1440)

165 KB PNG

>>109554091
Not trolling, why would I? Am interested in hearing people's opinion.

Always wanted to try it. Seems like a great way to learn. But I don’t know enough about assembly to get started. And I don’t know enough about OS dev to have interesting ideas to implement. I’d just end up copying some example, which isn’t inspiring.

>>109554360
not op and not even a programmer, but i do know about osdev... the website http://wiki.osdev.org/

>>109552794
they warned me because of this word
janny thread

>>109554360
if you want to just dick around in kernel without having to write the entire system from scratch, you can try xv6 or MINIX 2.
xv6 is super minimal and barebones UNIX v6 implementation, relies on cross-compilation.
MINIX 2 is much more complete POSIX compliant system (the first POSIX, before it got very bloated) with userspace full of useful tools (vi/make/less clones, manpages etc.) but it's kernel is still very simple and easy to read/edit. It also comes with a C compiler, so the whole OS is self-hostable, you can easily work entirely inside of it.
It also has a book accompanying it (the 2nd edition, not 3rd!).
Don't bother with MINIX 3, Tanenbaum shifted to do a more commercial reliable microkernel, bolted the entire NetBSD userspace on top of it, it's no longer focused on education, but he kept calling it MINIX for brand recognision.

>>109554429
I’m aware of it. But I haven’t found it to be much help.

>>109554637
>MINIX
Is it even possible to get something that old to run in the current year? It may be simpler than a modern OS, but I still wouldn’t have any idea what to do with it.

>>109552794
Its not. You asked AI do this. Its obvious.

>>109555139
Is it even possible to get something that old to run in the current year?
it's 32bit x86, it just works. I forgot to mention, but MINIX 2 has working networking and TCP/IP stack as well, i did my MINIX tinkering with it in QEMU over rsh/rlogin (predecessor of ssh, comes preinstalled), even wrote a custom network filesystem server for it (this way i could just mount the whole MINIX filesystem on host while it was running, a la NFS or SSHFS).
If you would like to run it on real hardware, it should work on most IBM PCs, for networking you'd need a network card based on one of those chips: DP8390, WD80x3, NE1000/NE2000, 3c503, RTL8139, AMD LANCE, Intel PRO.

my vibecoded os is progressing slowly
the kernel is running on a pi400 already

>>109555458
>>109554132
very nice, anons, tell us more about it.

>>109552794
>.tsx

>>109555555
checked

>>109555498
its based off a webos i built which i had codex convert to a plan for bare metal.
micro-kernel inspired
everything sandboxed, permissions need to be granted for anything applications want to do.
haven't decided on a filesystem yet, on my webos i had a fancy custom binary file to do everything.

>programs begin with no ambient authority
>useful access arrives as explicit endpoints and capabilities
>applications are packages with inspectable intent and provenance
>components communicate with named, version protocols
>large objects can move between components without becoming anonymous byte streams or unnecessary copies
>system activity, authority, failures, and component relationships are visible to the user
>rich services can be replaced without redefining the kernel
a bunch of other things, nothing new on its own i think, but many os ideas bundled together.

just picking high level ideas and letting codex run it in chunks and push to my pi400 for testing, using an esp32 that pretends to be a usb drive, collects uart and reboots the pi, so codex just hammers away for a few hours until it works on the hardware

File: IMG_4316.jpg (2.56 MB, 4032x3024)

2.56 MB JPG

>>109552794
>>109553641
> sidetracked

Haha - I’m doing the hobby OS thing on hard mode by first designing and implementing my own 8+ bit CPU (with custom ISA) and mid80s level graphics system and implementing them on FPGA I wrote my own assembler in C for it. I also want to make a sound chip in the same FPGA and hook it up to keyboard, mouse/joysticks, and SD card that is on the dev board for mass storage. I’ll do kernel functions in ROM like in C64 or PC BIOS, then a user shell with built in BASIC and maybe FORTH and eventually a simple GUI like GEOS or GEM. It’s been fun so far, but progress is slow oftentimes.

>>109555872
wow
way to fucking show everyone up anon
i kneel

File: IMG_4422.jpg (2.94 MB, 4032x3024)

2.94 MB JPG

>>109555872
Graphics display on VGA monitor. Some resolution modes are full 640x480 but I only allow 64k for graphics so I’m limited to character-cell color attributes or lower res modes like 320x480, 320x400, 320x200, and 160x200. I can do 320x200 at 8bpp but the 256 possible system colors are fixed out of a 15 bit color space.

File: IMG_4379.jpg (3.92 MB, 4032x3024)

3.92 MB JPG

>>109555895
Thanks anon, it is based on decades of programming and dabbling in electronics since the late 70s. I didn’t know VHDL for designing with FPGA until 10 years ago.

It’s not always smooth progress. I get a lot of garbage at times. :D

>>109555217
You're right to push back.

>>109555872
>>109555910
Incredibly based.
Did you ever find yourself going back to tweak the CPU when you're working on the OS and wished the CPU had certain features? Or are you keeping the CPU more set in stone and working with whatever limitation you ended up with post-design?

>>109553663
>Sony Vaio P netbook
Yeah, that's nice and all, but I'm going to take a wild guess that there's no GPU acceleration because lol GMA500

>>109557263
If i'm ever going to use a LLM for something regarding my OS (spoiler, I won't), it will be for reverse-engineering that shitty PowerVR-licensed core. But I ain't giving up, I have some NDA internal datasheets from Imagination Tech, and when I have time I'll dig more into them. Besides, Linux has a shitty driver too for that, and i think it's in a state between crap and shit, so it will be fun to work with.
But I don't really need hardware acceleration, modesetting would be nice though. Oh, and other goodies, like HDMI output for netbook similar to the Vaio P, but which have HDMI output.

Should i switch to qemu for testing?
I have just been using essentially an sd card multiplexer to test my pi os

File: IMG_4354.jpg (2.66 MB, 4032x3024)

2.66 MB JPG

>>109557246
Thanks. I’ve not really gotten to the OS level yet, just some memory test and boot message in the boot ROM and some other CPU and graphics tests. I did write a little routine that unpacks graphics images from a bigger binary blob in fast sequence to be a video player for very short, crude video clips.

Actually, after writing just a little code for it, I am itching to revise and reorganize the ISA and opcodes. But then I’ll need to redo the assembler. But I think that the new ISA will be something I’d be fairly happy with for writing code for the kernel functions, OS shell, initial language(s), and fun demos, games and apps. But implementing all the test code to validate all the instructions was very tedious and grueling the first time, so I’m really debating with myself on that. Maybe I can get a lot more clever about the validation techniques and do them much more quickly.

>>109557325
For my hobby computer I’ve mostly used AI only to look up weird issues with the FPGA tools I’ve gotten stuck on.

I did have Claude try to write a USB keyboard host driver for the ARM core that’s already baked into the FPGA/SoC… But that failed to work and all the crazy-deep troubleshooting it suggested (which actually seemed reasonable) could not resolve it either. Not sure the USB port on this dev board is implemented correctly or might just be super tricky to use.

Maybe I’ll use AI to port some well-known games to my system, but the coding can be the fun part, so why let the AI have most of the fun? And no AI will be pre-trained on my unique ISA, so then it has to depend on holding my specific requirements in its context window and I bet it would make a lot of mistakes.

there is no future for schizo projects, good that they are indistingushable from vibe coded slop these days, a serious proof that you shouldnt spend time on them.

beside no links provided to anything, it will disappear as tears in the rain, hehehe. compare it to T.Davis who publicly streamed and constantly pushed his thing around with his identity.

and ye, youre faggot, not a guy

>>109553997
What's wrong with POSIX?

>>109559750
Shut the fuck up you hobbyless golem, why do you assume people need to prove something? Most "schizo projects" is people scratching an itch or interest they have, there is such a thing as intrinsic motivation. Photographs did not make painting less enjoyable. The entire point is the process of making it. AI seems to bring out the best and worst in people, what >>109559733 is doing is not "slop", its actually really cool, but it also brings out stupid doomers(you) and yes slop/garbage made by boring people driven by extrinsic motivations.

>>109559758
everything

>>109553641
Literally me. Forth backend with opendylan frontend. Currently just about self-hosted. Have to do some renovation work on the compilation though.

>>109553997
This.
Fucking sick of this shit. Why are people like this.

>>109554056
It's sad, but all software announced these days will be highly scrutinized like this now.

>>109553917
Porting wine or proton is a massive endeavor. "Porting" chrom* is easier (you're not going to port anything, rather you will create a linux compat layer and run chrom* in there, same with other key software that works on linux). It's not super easy eitherway, but tractable.

>>109555872
B-B-B-BASED OVERLOAD!

>>109559733
My experience with AI tools for similar work but in much more common language pairs (e.g. js, ts, python) has been disastrous. Basically, it always does something that looks right if you only look at input and output (the more complex the more "it looks right" because the harder it is to actually verify if it's right) and then it's totally wrong. You just read the code and it's not even close to what you instructed. If you point it out it will try to create tests and fix it and inform you it's totally passing all tests now. You check, it's even more broken than before (it's just hacking around the tests). If you give it tests, same thing, it will just hack around them and never make progress.

Maybe in a decade.

>>109559758
Everything

>>109554056
in your fucking dreams nigger, most obvious vibecode fonts and ui of all time

>>109560483
what harness/model did you use?
hasn't been my case at all

>>109560720
I tried just about every meme around. Codex, claude code, pi, roo, opencode. I tried glm and deepseek models, as well as maxed out anthropic and openai models, the latest of which would have been opus 5 (fable doesn't work due to (((safeguards)))) and gpt-5.6.
I also tried gemini once or twice but found it was always total garbage.

I customized a series of skills and a prompt library to max out the outcome and compared my workflow to others in the industry. I also exchange routinely on the subject with my coworkers, but they have the same experience I do with these tools. Digression: our boss wanted to hire someone who's great at using AI tools to 'teach us' how to use them right and get faster. 100% of applicants were pure vibecoders with 0 harness or anything and their code was pure slop despite instructions clearly saying code quality is paramount.

Eitherway, I also tried these harnesses and various setups with various skillsets and techniques in several scenarios, from very simple webslop to more technically advanced topics like compilers, practical time-saving applications like reverse engineering, decompiling and translating code in one language to another language, and to generate glue code (bash and python scripts that just interoperate off-the-shelf software).

At best, it sometimes manages to do a few things right with webshit, but even then it's limited (like I have to manually scaffold the project before it can actually do anything useful without hallucinating package names and versions that don't exist all over the place). After that, I can ask it to change a button's color or to align a div, but anything more than that and the code quality is trash, even though I use a self-judgement loop.

I also make sure to use minimum thinking budget for classification-style tasks in the pipelines/workflows I use, because I built a gold set to evaluate perf and found it increases perf by 15 points on average.

>thread about a genuinely cool topic
>gets derailed by AIfaggots
Every fucking time.
Ban them all. Give them their own site or board or something. Fuck, they just get more annoying by the day.

>>109560505
Font is standard Arial.ttf. Well done.

>>109552794
What is your OS's KILLER FEATURE?

>>109560877
If i had to choose a single one?

Networking: full TCP, IP, DNS, DHCP, IPv4 (fuck IPv6), SMTP, POP3, NTP (for timesync), HTTP, HTTPS, FTP, Gemini, Spartan, Gopher, Finger, Ident (for funsies), SOCKS4/5 and Tor (obviously not full protocol implementations but the bits I need to securely connect to servers and get what data I need).

>>109560895
I don't see that as a killer feature, it's just pretty standard and available on any other OS. A killer feature needs to be something your OS at least does differently, not to necessarily say 'a lot better' (the reference figure often used is "10x better").

>>109561606
I see your point but it's a one-man job in 6 years, it's hard to pinpoint something like "well now i can connect to Voyager and issue orders" because most of the dark years i've been stuck into bootloaders, and semaphores, and mutexes and deadlocks.
I'm still pretty proud of my OS's Tor support, my OS is the only hobby operating system in the world which has that. At least to my knowledge, it's been a few years since i last visited that assholes-filled shithole that we proudly call osdev.org.

>>109552794
based but i could never

>>109561657
That's absolutely not an excuse. TempleOS had several key features, not least of which Holy C. Not saying those were marvelous or anything, but they were there. Originality has very little connection to manpower, budget, or time spent: those affect only execution.
For example, there is no reason to be proud of your tor support: any other hobby os that has a unix compat layer can run tor without having to reimplement it, and is tor access really the thing that makes an os better/more interesting? It's a tiny userland concern that is only relevant in a super niche case, so other hobby os's not supporting tor is actually a good thing: it shows they're deprioritizing irrelevant concerns.

You did a lot so you should be proud of that hard work, but this and that are very different concerns. You could, for example, work hard at hand-typing the bible from your reference kjv, but that would be equally wheel-spinning when you can copy the already transcribed text anywhere.

>>109561685
You cannot downplay Tor to "anyone can do that, provided they have a TLS library embedded, and a full network stack, and they understand the ntor key exchange, and they have a full crypto library available, and they understand how EXTEND2 works over multiple relays, and they understand why a tor relay is totally within its rights to do absolutely nothing to your request, and you spend 2 days debugging absolutely nothing".

But yes, everybody can do everything. Point is nobody has it.

As for the second part of your message, I won't even bother to answer. Whatever makes you happy.

>>109561727
That is not a downplay. You are greatly aggrandizing the scope. First you say 'b-but I have tor' which I rightfully pointed out takes virtually no work for anyone else to have, and is about a last priority for all but anonymity-first OSs (which yours isn't).
Now you're saying "b-b-but to have tor you need a net stack". But that is a completely different topic, and most hobby OSs get a net stack very soon. Anyone can get any tls library from anywhere, they're a dime a dozen and don't take any effort to port, either. Basically, you are saying that your house is the best because it is painted red and you think other hobby OSs use other colors (and then you defend yourself by saying that you can't paint a house without walls). That's nice, but anyone can just repaint (and everyone has walls). So it's not a killer feature, simple as.

>>109561771
Nta but LOL: only on 4chud can people claim an almost complete (from what the anon said) network stack is something trivial.
>and most hobby OSs get a net stack very soon
HAHA

>>109562007
One would have to be very special in the head to actually believe you are
>Nta

>>109552794
Release it faggot

>>109552794
Post repo or didn't happen.

>>109560462
You're right. idk OS dev beyond extreme Linux or BSD tinkering. And it's hard to trust shit when AI can produce a somewhat professional looking interface while the actual code is abhorrent.

>>109559889
>its actually really cool

nope, i though it was nice a while ago, i even re-posted some DuckOS meme, but now ive changed my mind. these projects are either copycat retardation or LARPing of High-IQ persona (mentalist cosplay is antient circus). there is no other purpose. but this particular LARP is generated because as i said

>compare it to T.Davis who publicly streamed and constantly pushed his thing around with his identity

this simply natural for anyone who put lots of effort into the thing, it must have a history and stuff. youve been tricked

>>109553301
>*starts running with him*

File: why are you gay.png (122 KB, 480x360)

122 KB PNG

>>109552794

> /g/ays, why is it

>>109555555
CHECKED
>>109552794
>>109553663
>>109553663
>>109560895
Very, very cool, anon. Mad respect.

>>109552794
Physics girl!

>>109552794
give us a download