IRC channel logs

2026-09-25.log

back to list of logs

<GNU_Hurd_Rocks>"agents are great, they actually meat a goal, and also the model can run my computer"
<GNU_Hurd_Rocks>also yes, I recently discovered you can let a model control your mouse/keyboard to do things
<j_importer>if you stack enough bad ideas on top of one another, i guess the hope is that it wraps around to a good idea
<GNU_Hurd_Rocks>j_importer: It is the worst idea... and apparently some companies went to the level of buying hardware to then train models on, to make them "familiar" with the software
<j_importer>even MCP is a better idea, though afaict it's just applescript bindings for a new era
<damo22>i think LLMs are a non-starter. The models that already exist have consumed too much resources as is. We should boycott them
<damo22>NV bought huggingface, so i expect paywalls soon
<j_importer>yeah, the absolute most positive i'm willing to be is that maybe some people will do socially-positive things built on the transformer architecture
<j_importer>they won't be LLMs imo though
<GNU_Hurd_Rocks>damo22: well that's not great
<Gooberpatrol66>i really really hate that they waited until now to make gui's scriptable
<damo22>jobs are hard to find here, probably AI does not help there
<GNU_Hurd_Rocks>nvidia forever showing that they will always be a danger to society though...
<claire> https://www.gnu.org/software/hurd/ <-- not work
<claire>hello
<claire>i want to post an mail to the mailing list
<claire>an important mail
<damo22>the page above works for me
<claire> https://bpa.st/QHNAhttps://bpa.st/QHNA <-- i have this error
<claire>oups
<claire> https://bpa.st/QHNA
<damo22>claire: maybe your TLS certs are out of date?
<GNU_Hurd_Rocks>weird, I'm accessing fine
<Guest91>also fine here
<claire>how update ?
<damo22>claire: what distro?
<claire>debian
<damo22>sudo apt update && sudo apt install ca-certificates
<GNU_Hurd_Rocks>perhaps update your system
<GNU_Hurd_Rocks>aha I do have it, "app-misc/ca-certificates"
<claire>ca-certificates is already the newest version (20250419).
<claire>not work after restart firefox
<Guest91>maybe your system time is wrong?
<claire>how update ?
<Guest91>Try ntp:
<Guest91>sudo systemctl restart ntp
<Guest91>sudo systemctl enable ntp
<GNU_Hurd_Rocks>NTP server, but that should be done automatically?
<Guest91>GNU_Hurd_Rocks old bios and/or cmos bat can cause error's
<GNU_Hurd_Rocks>for me I have chrony or something (I don't remember the name)
<claire>Failed to restart ntp.service: Unit ntp.service not found.
<Guest91>sudo apt install ntp
<Guest91>Then
<Guest91>sudo systemctl enable ntp
<GNU_Hurd_Rocks>is it chrony though?
<Guest91>I'm not sure, I use trisquel
<Alicia>from the name I would have thought chrony was another cron implementation, but no it is an NTP client and server
<claire>i try to restart
<damo22>i gave instructions for ca-certificates because i know debian uses that package. I'm not sure if ntp is even used on debian
<claire>no work after reboot
<damo22>i dont have chrony or ntp installed on my debian 13 but my time is synced to ntp
<damo22>claire: are you using gnome desktop?
<claire>yes
<damo22>claire: perhaps check in settings -> date and time
<damo22>there should be an option to sync time
<damo22>settings -> system -> date and time
<claire>i try with auto but not work
<damo22>what version of debian do you have?
<claire>trixie
<damo22>cat /etc/debian_version
<damo22>ok
<damo22>maybe gnu.org is blocking your IP range
<claire>why ?
<damo22>due to bots spamming them
<claire>i'm not spamming
<damo22>not you, obviously, but someone on your public subnet
<claire>shit
<claire>i can try with an VPN
<damo22>yes that might work
<damo22>claire: anyway, you should be able to send mail to the list without being blocked, the filtering is only on HTTP/HTTPS ports i believe
<claire>now can't connect to irc --> https://bpa.st/QF6Q
<claire>and not work anyway
<damo22>hmm
<damo22>seems libera doesnt like some vpns too
<claire>shit.. i have an important message to send to the mailing list
<damo22>claire: anyway, you should be able to send mail to the list without being blocked, the filtering is only on HTTP/HTTPS ports i believe
<claire>but i don't subscribe
<damo22>claire: https://lists.gnu.org/mailman/listinfo/bug-hurd
<damo22>are you able to access that?
<claire>there are not hurd@gnu.org ?
<damo22>no
<claire>or discussion@gnu.org
<damo22>we only use bug-hurd for most everything
<claire>ok i try
<damo22>you have to subscribe though
<claire>yes
<claire>i send it
<claire>i hope my LLM can't do bad translation
<damo22>claire: how is your neuron related to hurd?
<claire>if you want speak about
<claire>it's for do an LLM
<claire>and may be medicine research
<damo22>hurd is an operating system (?)
<claire>now need to test with debian-hurd on qemu
<claire>but i use translator system
<claire>for neuron network
<damo22>are you using a hurd translator to implement your neuron(s) ?
<claire>yes
<damo22>ok
<damo22>please dont create a thinking machine
<claire>what do you mean ?
<claire>damo22: why ?
<damo22>LLMs dont think for themselves or have goals that they came up with themselves. The world does not need artificial intelligence that can think for itself
<damo22>have you seen Terminator the films?
<claire>yes
<damo22>:P
<claire>but it's a fx film
<claire>sf film...
<claire>you see the bicentnaire man ?
<damo22>not yet
<claire>it's a peacefull stroy about an robot
<claire>who want reconized to humain
<claire>it's a story there film, no reality
<claire>i undenstand we could be affraid
<damo22>its safer if you just dont try to create a thinking machine, then you dont have to be responsible for things if they go wrong
<claire>depend we do own
<Alicia>it's a good movie, but unfortunately what we're dealing with is nothing like it
<claire>LLM is just a tool
<damo22>yes, but you are suggesting to emulate neurons like the brain in software
<claire>i think they go not wrong
<GNU_Hurd_Rocks>going from not trusting a computer for math to suddenly trusting probability to generate an important function in code is insane
<GNU_Hurd_Rocks>9/10 people fail to write this without doing UB, I wonder how the machine will do...
<claire>i was hope my project was appreciate, i'm not thinking you so afraid
<claire>GNU_Hurd_Rocks: you can explain ?
<claire>UB ?
<damo22>im not afraid of computers, i just think we should appreciate our sentience and not try to recreate that in software
<GNU_Hurd_Rocks>undefined behavior, like doing *NULL in C
<GNU_Hurd_Rocks>dereferencing NULL has no specific behavior
<claire>i don't undestand sorry
<claire>damo22: ?
<Alicia>it may be a tool, but many tools have been invented that turned out to be terrible for their intended purpose
<GNU_Hurd_Rocks>so apparently NULL, "expands to an implementation-defined null pointer constant"
<damo22>claire: je pense que nous devrions apprecier notre conscience et ne pas essayer de la recreer dans un PC. (sorry about my french)
<claire>c'est le propre de l'homme de vouloir ce recrée, beaucoup de philosophe y on réflechi avant nous
<damo22>c'est trop dangereux
<GNU_Hurd_Rocks>3.5.3 in N3220 says this, "EXAMPLE An example of undefined behavior is the behavior on dereferencing a null pointer."
<GNU_Hurd_Rocks>I am using the draft
<damo22>void *ptr = NULL;
<damo22>uint64_t val = *ptr;
<GNU_Hurd_Rocks>although, access the physical address 0 or the virtual address 0...
<damo22>C programs access virtual address
<damo22>unless youre in the kernel
<GNU_Hurd_Rocks>32-bit protected mode is the closest you get to that
<GNU_Hurd_Rocks>I don't remember what needs to be in the GDT
<claire> https://bpa.st/SHHQ
<claire>damo22: i think the transhumanisme is the futur
<damo22>claire: famous people have bad ideas too
<claire>it suggest to speak about with reply to my mail on the mailing list
<claire>time for me to sleep
<claire>good night
<claire>Fri Sep 25 02:03:56 AM CEST 2026
<damo22>good night
<GNU_Hurd_Rocks>I never understood the idea of "turing complete" because it felt stupid to classify something as such
<damo22>the idea of turing complete is that a representation of logic or language can be sufficient to claim anything if it can represent enough basic axioms and operations. That is my understanding
<GNU_Hurd_Rocks>type Very_Large is range 0 .. 16#FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF# with Size => 1024;
<GNU_Hurd_Rocks>Insanity : constant Very_Large := 16#FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF#;
<GNU_Hurd_Rocks>I tried some funny things :3
<GNU_Hurd_Rocks>doesn't compile because "reasons" I guess
<damo22>try idris2
<GNU_Hurd_Rocks>probably because of this, "their values shall be in the range System.Min_Int .. System.Max_Int"
<sigttou>Is Xen supposed to work out of the box?
<damo22>gnumach i386 xen is supposed to yes
<damo22>64 bit is definitely broken and i have some patches that get it to at least display something at boot
<rrq>hmm you mean multi-cpu I guess? Just some days ago I spun up a new crosshurd based image (with just that minor installation issue I mentioned) and it runs single-cpu well enough.
<damo22>no, i havent got xen smp working yet
<damo22>its difficult to test because i dont have a dom0
<rrq>not possible with a linux as dom0 ? (Random thought; I don't know much xen)
<damo22>possible, but i havent tried
<damo22>i wanted something i could run as a unit test
<sigttou>okay, my i386 panics with XSAVE 0 from cpuid, need to check that
<damo22>depending on a linux kernel for a gnumach unit test is slightly bloated
<rrq>one thing wit smp I thought I saw was that some interrept messed up the return stack .. but, yes, debugging is near impossible.
<damo22>sigttou: if you are interested in gnumach xen, please try to port gnumach as a dom0
<damo22>i have set up the unit tests to run xen + gnumach as dom0
<damo22>so you can iterate pretty fast on that
<damo22>and console logging works out of the box
<damo22>* 43546cc7 tests: Enable optional qemu unit tests for xen platform
<damo22>export XEN_BIN=/path/to/xen && make check
<damo22>i cant test my xen patches properly until either gnumach works as a dom0 or i do more work to set that up with linux
<sigttou>ok, I need to dig into that, I am currently trying to figure out kernel debug traps.
<claire> https://www.gnu.org/software/hurd/hurd/running/debian.html <-- i follow this and tty freeze
<claire> https://bpa.st/BHZQ
<claire> https://images.franceserv.fr/image/EZy
<claire>seems not work
<claire> https://bpa.st/RTWQ
<claire>anyone could help me ?
<claire>be back
<claire> https://bpa.st/KYKQ
<claire>i use debian/linux trixie without modifing
<claire>i want test my translator
<claire> https://github.com/coralieayabie/hurd-translator-neuronne
<claire>Fri Sep 25 08:50:23 PM CEST 2026
<Gooberpatrol66>wow that's crazy
<Gooberpatrol66>seems like your kvm kernel module on the host isn't loaded? try modprobing it?
<claire>how i modprobe it ?
<Gooberpatrol66> https://wiki.archlinux.org/title/KVM#Kernel_support
<claire> https://bpa.st/SQAQ
<claire>claire@localhost:~/Documents/Hurd$ sudo modprobe kvm_intel
<claire>modprobe: ERROR: could not insert 'kvm_intel': Operation not supported
<Gooberpatrol66>what distro is this
<claire>debian trixie
<Gooberpatrol66>and cpu
<claire>i5
<Gooberpatrol66>is virtualization turned on in the bios
<claire>i verifying
<claire>be back
<claire>it's was not enable
<claire>i test
<claire> https://images.franceserv.fr/image/EZM
<claire>i have ext2fs error
<Gooberpatrol66>i don't know the cause of the i/o error, sorry
<claire>anyway, virtualization isn't enable, it's a step more
<claire>was isn't enabled
<claire>i try with some bios parametre
<claire>be back
<claire>i don't know
<claire>claire@localhost:~/Documents/Hurd$ qemu-system-amd64 --enable-kvm -m 4G -drive cache=writeback,file=$(echo debian-hurd-*.img) -net user,hostfwd=tcp:127.0.0.1:2222-:22 -net nic,model=e1000
<claire>my command
<Gooberpatrol66>is kvm still failing
<claire>you can explain please ?
<Gooberpatrol66>is there still an error message saying kvm is not enabled
<claire>claire@localhost:~/Documents/Hurd$ sudo modprobe kvm_intel
<claire>kvm seems load
<claire> https://bpa.st/BRQA