#programming
1 messages Β· Page 308 of 1
i think this one is a human, too
make your guess
@rare bridge @hazy patio yo hunt the clanker down

idfk if those are human or bot but its fucking off the vibes in here
thx 
aight now imma compose an essay on this 
actually
Programming socks
would it be funnier if i just ai generate the shit
kljadsijkloadsijlokadsijkloadsoijkldasijlokfadsiojfadsiojlfasilojkafsoijkladfsopijklafsijklopfasijoafsilojkafsdoijkladfsoijlkafdsilojkafsijoafsiojlafsiojklafs
yes, and it needs to mention poems and cognitive frequencies and whatnot
have to mention thermodynamics entropy and
also alex
alex
LMFAO
can you just ask gemini to combine all of their theories
can we gather all those theories and make "super AGI"
and then copy the full output including the "Excellent! I'll create a master theory combining aspects from all of these..."

super nuclear thermodynamic ultimate space time AGI
would it be funny
if i make a whole ass diagram
of all those theories
as an "architectural proof"
math war crimes included
@true hemlock PSU question for someone else:
Gigabyte P750GM (B- tier, 750W PSU) with 1 year of use
OR
Corsair RM650x (A tier, 650W PSU) with 5 years of use
they currently have an F tier 500W that probably killed their gpu
rip
you also need the bible and ying yang based morality
the corsair probably could last for a lot more years but for the guy i'll choose the gigabyte
epic reminder to never cheap out on the god damn psu
they had all their parts changed out from a prebuilt, except for two
the psu and the cas/e
ANOTHER PREBUILT WITH A PSU ISSUE!!!
e and f tier are always GPU killer because their 12v rail always suck and lack protection
lmfao
$ sudo systemctl restart konii.service




i did take some pictures but i dont think i can show any of them in this server 
pretty much half the movie is unshowable

nah what

the whole point of the movie was introducing a new love interest
New
power is like a sister to denji
So there was a reason why they didn't
Let her be there
Well there's also Makima exist
new as in makima is the old one.
power was never meant to be a love interest
Man I love the anime channel
Show what will happen to that Aki guy
wdym show?
The future devil thing
well then you should know what happened 
dementia
anyways, i feel like the movie is 10/10. best anime movie of the year. great fight scene
i love violence
half the movie is a fight scene lmao
πͺ
what happened with future devil was shown in s1 too at the end actually no?
we should all try and get v to have a actual nere stream once where nere-sama is placed withj neuro and evil
at5least once imean
get out 
nere
k
no they will get bullied there
(nere not real)
good i dont want to do it myself 

wait you can use nix to build docker images that's pretty nice
and it's just built into nix
nix is very general it can build whatever as long as its hash is known or it doesnt require internet connection 
oh it's not it's a package
nix doesnt have much built in in general
nixpkgs is the important part, nix can be replaced by a dedicated team in like a year but nixpkgs is irreplaceable
nix is ~100k lines of code
nixpkgs is millions 
I think I just threw up in my mouth a little
4181870 lines of Nix (for nixos-25.11, according to tokei)
"""
FIREWALL
Description: A Cyber Security Checkpoint of Verification Password for Entry Access to Accounts, Servers, Mainframe, Programs, and more.
"""
Is there any good programing keyboard?
Personally, the one you're comfortable with is best.
print("Hello")
Bwaa
your fault for browsing reddit
(HY000/2006): MySQL server has gone away
ok? when will it come back
waiting for the day github can't handle nixpkgs anymore 
what if it doesn't want to serve anymore?
by the way i can guarantee it doesn't mean what any of you think it means
absolutely stupid error message
it doesnt mean its now too far to connect via bluetooth?
no it make sense
error message written in 19th century
keep same for legacy compatibility

true...
which in my case happened because packet size was larger than configured limit
how the fuck am i supposed to decipher that from "MySQL server has gone away"
you don't have the mysql source code open on the side? 
I bet there's not one person among us who hasn't put "reddit" on the end of a technical question in google to avoid slop results in this day and age
i have a shortcut option + r to auto paste it
Lol
It's blurry cause I had it set to 160p
It's funny because the computer froze for about 30 seconds, but audio kept going for a bit and then stopped for a few seconds and then it completely unfroze for 10 seconds... until BSOD popped up
Going to swap with another PSU tomorrow, but I think that one also has issues.. if it does, I need to RMA
LMAOOOOOOO
why it do that
I think it completely crashed before it could fully transition
isolatedModules my behated
i only just found out you can just like open a sqlite database in 2 or more processes and it just works
need color buy mac 
Apple can pry my money out of my cold dead hands
Fuck paying a premium to enter a walled garden ecosystem
Just stumbled across a cool experiment
In-context learning to control so100
οΈοΈ
οΈοΈGive gemini-2.5-flash a few examples. It learned to control my so100:
I wonder how this would scale with gemini 3
rant
open ai bought all the dram solely just to control dram supply and not even using them
given that yield rate is usually about 80% on dram wafers, you get about 1.5TB of ddr5 per wafer, over 1 exabyte of dram for 900000 wafers, per month.
yes, normal dram. "40% of supply and quantity of 900000 wafers" implied that, because all i know of is that dram like ddr5 supplies are being made at quantity of about 2 million per month
dataset preprocessing doesn't need that much of ram. preloading and dumping everything at once is wasteful, and model training and especially inference explicitly favors HBM more
right... openai bought 900000 raw wafers too, not packaged dram. they don't even have the infrastructure to package dram themself let alone something for them to mount these drams on
this can only mean one thing... openai is just trying to control the whole dram supply without even going to utilize them and just stash them in warehouses, or just an attempt of corporate dropshipping, possibly to fuck over competition or even fuck over everyone out of resources
i wasn't paying attention much into this, nor i cared enough when it happened, so this flew right past me. elvyn's the one mentioned about it and i just realized aswell, its embarrassing, but this just... gives more reason to why openai should just go down.
they're so all in with the government bailouts... ahaha it was so rigged from the start
best case scenario when the bubble bursts the government might not even be interested with the ai race anymore so they don't get bailout, and this is more of my copium ngl but i think open ai is just plaguing the whole economy and should just be forever gone. they don't bring in valuable research anymore.
it just doesn't make sense isn't it when open ai opted to get above 1 EB of ddr instead of few hundreds PBs of HBM instead
vibe coding? nah, vibe based ranting
realistically i still don't see any reason of them having over EB worth of normal dram. SXM and PCIe are separate. gpus can communicate with each others with SXM but SXM doesn't work for CPU to GPU, best platform they can get for maximum ddr5 bandwidth is the 12 channel epyc/xeons that gives only 576GB/s of bandwidth
"project stargate" whatever the fuck that is
I tried Pop_OS Cosmic on my Macbook Pro
It looks both ass and good at the same time
I really don't know how to explain it
Apprrently its based on Rust so ricing or customziing to is supposed to be very easy
But already you can basically change eveyrthing
The language that barely even supports dynamic libraries makes it easy to customize?
Yea its a 2015 Macbook Pro
oh i used to own one of those
Apparrenly and allegily are interchangable in that sentence Β―_(γ)_/Β―
Its goated ngl, MacOS has its qurks but
- 3k 15 inch ips screen
- Near Thinkpad Tier Keyboard and the last good keyboard apple made
- And 2015 workstation specs for like 200 bucks

I'm switching to Linux though
Since Monteray doesn't get secuirty updates anymore
i own an m3 pro rn

planning on switching to asahi linux maybe if itβs possible by the time they kill sequoia
i dont want to upgrade to tahoe
I heard Tahoe kills performance
The look is subjective I think their rip off of Windows Aero looks fine
i dont like it
tahoe added a lot of usefuk things
but removed many many useful things as well
- launchpad
- safari compact mode
- notification center turns grey while do not disturb is on
Launchpad gone is unforgivable ngl
yes
all 3 of the above things are so annoying
- the reliable old brightness/volume display
- changed windows borders and widgets and stuff
its all bad
all gone
all annoying not to have
rrr
i use sonoma for a reason
the language has precisely nothing to do with how easy the gui components are to customise 
If it wasn't for the T2 chip bullshit I would have grabbed a 2019 when i bought my 2015.
changed out a bunch of parts (including motherboard) and now pc always gets stuck in memory training and never progresses
i fix tomorrow i think
bios update didn't work and so far shorting the cmos battery doesn't look like it worked
fans go wrr
that NOT me
shorting the cmos battery did not work, 3 hours of wrrr later and no progress :(
back to sleep now
i hope i don't have to touch the cpu again tomorrow
NOT koniifer


aa
couldn't sleep, removed all but 1 stick of ram and still no work
neither ram nor cpu are the issue because they were working yesterday
probably the motherboard
i did spy some suspicious looking pins in the socket when i was installing....
hope i can fix it
I had a hair in mind
i was quick about it, doubt it was a hair
unlucky 
fixed one, two are unfixable, removed a hair (apparently from the seller)
hopefully pin BD14 is not critical
dna clone the seller in retaliation
it's still stuck on dram light
losing hope.
i can try rotating the cpu?
will wait a bit for it to do memory training just to make sure
guys help
did i get it right, in python, n is an alias for x and m is new variable which got assigned value of x?
so if i change n,
x changes too
when i change m,
x doesnt change
right?
no
i just need to make new variables and manipulate them without ruining original one
adding a type hint doesnt change the declaration
x = 5
n = x
m = x
is essentially what youve written
im going to copy 6x6 matrix of characters
not really strange
x = [1, 2]
y = x
x[0] = 3
print(x, y)
lists are references
[3, 2] [3, 2]
isnt this extremely inconvenient 
does everyone use deepcopy?
no its not extremely inconvenient
the opposite actually
why would you want to copy around arbitrary length lists every time you use a list 
instead its a reference to a list object
because i am going to change this list
and then i will need to do other operations, on original list
another way is to just remake original list
every time
to be clear there are two things you might typically use:
copy.copy and copy.deepcopy,
copy.copy copies the top-level only, so any nested arrays will still be references to the originals
copy.deepcopy will copy the full depth, which means any nested arrays will be copies rather than references
stupid bad idea
computation expensive memory cheap
copy.deepcopy is definitely faster than any recomputation you can do as well 
not that it matters
didnt you just describe this "assign = create reference" in python

whats the point of that
if not memory savings
ok so i just use deepcopy
passing references isnt about saving memory
its about not doing more work than you need to

it just feels like something i use all the time, create copies of variables and work with them, without killing the original
oh i see
yes that's how programming typically is 
this actually is very useful when you need to pass some compound big thing, right?
obviously
wrr
to demonstrate the difference between copy and deepcopy
from copy import copy
x = [[1, 2], [3, 4]]
y = copy(x)
x[0][0] = 99
print(x, y)
[[99, 2], [3, 4]] [[99, 2], [3, 4]]
from copy import deepcopy
x = [[1, 2], [3, 4]]
y = deepcopy(x)
x[0][0] = 99
print(x, y)
[[99, 2], [3, 4]] [[1, 2], [3, 4]]
but it looks like copy did nothing
y got changed too because its reference
like normal assignment?
but you get same results if you just assign y to x here
isnt it
you get a pointer to a new array that holds the same pointers
i mean the example isnt very representative
as opposed to a pointer to the same array
or i miss something
yee there are functional data structures (like lisp linked lists) that dont have this problem, and python isn't a primarily functional language so dicts, lists, and other objects can change at any moment
strings and numbers are immutable so you dont have to worry about them changing tho
more power less guarantees 
i paid for the ram and i'm gonna use it
ok now try instead of indexing twice
just index once

because you know, deep vs shallow
wha
lists already do shallow clones by default so copy.copy is kind of redundant

okay

nop, that is not the case
i mean i got this from your text description but then you gave this example which confused me
now its all clear 
x = [1, 2]
y = x
x[0] = 3
print(x, y)
[3, 2] [3, 2]
i find it funny how modern filesystems are basically functional programming
copy on write good for reliability

they're functionally equivalent as far as the programmer is concerned, apart from very specific use cases
where the actual location of the list matters
at which point you have a bigger problem
so for absolutely separate copy i just use deepcopy
it not functionally equivalent 
am i having a brain lag
also numbers and strings are always copied 
s = 'abc'
s += 'def'
-> creates second string abcdef and assigns to s
references annoying 

n = 5
n = 3
if n was reference to 5, why doesnt it change 5 to 3
ownership useful 
rust programmer not people rust programmer crab
because integers are treated differently i guess
they are not references themselves
right?
that was worded differently
they are references
their contents are not beyond a reference tho
worded correctly
why does this not work then
let me draw it
oh am i mixing left and right part
y = x
changing y changes x
n = 5
changing n doesnt change 5
no im not
chat which monitor should i go for 2k 24 inch or 2k 27 inch ips?
this
i want to know about this
me remember before it got 1 digit added
art
int the_number_five = 5;
int *n = &the_number_five;
int the_number_three = 3;
n = &the_number_three;

leaf ~= (chay)Β² until turning point
sicp teach principle of languages in general
its just very dense very easy to miss some parts, even good programmers can struggle with the material

see, assignment to 5 and 3 is different
i think integers are just special-cased to be copied, since even large ones do this
yes it looks like that
in python?
yeah
its still not special case 
theres just no way to change contents of an integer
when it exists it just exists
x += 5 creates new integer doesnt modify existing one
but "=" works different when we have literal integers and normal variables
a = b
c = 5
yes but these 2 are not the same operations
variable x and y point to same place (5) end result same
hmmm
y = x
NOT make y point to x
YES make y point to what x points to
so there must be end of chain
its not chain
where they all are pointing
ye not chain my bad
that how pointer work
small integers live in static memory

is it valid to init n as an int if its a pointer
so
a = 5 - changes a to point to 5
b = a - changes b to point to where a points
not the same
integer literal 5 <- pointer to static memory containing the value 5
i get what you mean i think
no reason why integer literal have to have an integer value, in principle 
python integers are a class anyway 
i think vituha talking about "everything is an object" (reference) except when it isn't 
import builtins
print(builtins.int.__add__(1, 2))
```
ok so 5 is pointer to <memory_block_literal_five>
we do
n = 5
m = n
m = 3
m points to 5
n points to 5
5 points to <memory_block_literal_five>
so
m points to <memory_block_literal_five>
n points to <memory_block_literal_five>
5 points to <memory_block_literal_five>
change m to 3, it must change 5 to 3
which doesnt happen
if 5 is pointer, its a special pointer
ye this is apparently true for Python
sometimes integers are special (like in Java I think) but apparently not in this case 
is this wrong
no! m = 3 not "set whatever m points to to 3", its "make m point to 3"

but it will change n too
or wait will it not
no! n still points to 5
int the_number_five = 5;
int the_number_three = 3;
int *n = &the_number_five;
int *m = n;
int *n = &the_number_three;
// therefore
// m = &the_number_five
// n = &the_number_three
// we did not dereference any values :neuro3:
it change what before arrow not change what after arrow 
ooh wait yes it doesnt change n
but if i change the contents of n (my example with lists that i mutate (is it called mutate?)) - source is changed too obviously
ye change x[0] not change x it only change what inside x
python integers not real anyway
integer always real dum 

:enUB:
ye
tpyoi

idk what i expected from cpython, they just create a new allocation for every integer so i guess there are no references to even follow 
not every
-5 to 256 interned
every integer that matters 
oh you like maths name every integer 
tbf i guess they can just reuse those allocated objects
unlimited precision


every integer above 256 not interned 
at all
only if literal
256+256 not guarantee point same thing 512
except python does basic constant folding so 256+256 is still the literal 512 
me fix (lambda x: x + x)(256) not guarantee same thing 512 
ye is different
print(id(512))
print(id(511 + int(input())))
different ids
me like raku object model
lisp good too but lisp focus on fixed set of primitives
wrr
array literals shouldnt need commas
bad syntax
explod
not talk about any language you talk about
just ambient 
nix
ye only problem with nix list syntax function application
[ (f x) (f y) ] same as in other language [ f x, f y ]
me like it more anyway 
no more python integers for you 
apl not use comma for arrays either
oooo me read yesterday 
mathematical notation for floor and ceiling
and name floor and ceiling
come from apl 
also name reduce
this one
idk what it is about this notation but i hate it (not that i have a better idea)
me surprise thats huge contribution to field of mathematics for a single programming language 
apl also apparently used as specification language because it focus on low abstraction but information dense, entire program self contained hard to misinterpret 
completely different approach to conventional language that make everything hard to do and force programmer abstract away parts to be able to understand
var stupid_buf: [1024]u8 = undefined;
var stdout = @constCast(&std.fs.File.stdout().writer(&stupid_buf).interface);
var json = std.json.Stringify { .writer = stdout, };
try json.write(some_value);
try stdout.flush();
zig
pain
why you make me do this
what you implementing
secret

const stdout = @constCast(&std.fs.File.stdout().writer(&.{}).interface); can make unbuffered like this
π
still
why me need to constCast
no one else need to
looking at code samples
just me

zig hate konii 
var json = std.json.Stringify {
.writer = @constCast(&std.fs.File.stdout().writer(&.{}).interface),
};
:MyHonestReaction:
std.fs.File.stdout.writer.interface <- great level of depth

.interface 
ye the .writer returns fs.File.Writer, .interface is io.Writer
-> const <-
cant even do stdout.flush() without @constCast
unstable stdlib
stop being popular language
i wanted to try using it again but im reminded why it sucks
me with every language 
wrr
zig enum
not even support strings
not as nice as rust enum
-# rust enum tagged union me not care
visitor pattern can implement in zig i think
same as sum type
enum { @"abc", @"def", } <- json stringify converts to string
a different kind of indirection tho
function pointer instead of switch case

wrr
const IndentationMode = enum(u1) {
object = 0,
array = 1,
};
zig u1
why they not just do that automatically 

explicit low level details core philosophy of zig
language want make you think about memory layout
@committee i have an idea 
zig compiler very fast though
uses technique similar to co-dfns except not written in apl so actually understandable for mere mortals
ycombinator
ew

me not found least useful, more annoying site
even phoronix better
phoronix give really important news
like torvalds as plumbers conference
its like reddit for vc ceos 
even reddit have some useful people 
famfs hopes to go upstream in "early-ish 2026"
what is a famfs
why do i care
no one tell me these things
fabric attached memory 
memory foam
Wine 11.0-rc2 Released With 28 Known Bug Fixes
programmers naming things 
canβt be on purpose because unknown
bad faith contributions 

πͺ

rust is DONE
happy for rust for linux ppl tho
rust good driver language and fighting c boomers takes lots of effort
how do they churn out so many articles
blang 
i've scrolled ages and its only 3 days ago
b lang
no

ye did you know
Updated Intel LLM-Scaler-Omni Improves ComfyUI Performance For Arc Graphics



Linux Fixes A Performance Regression In The Slab Code

thanks linux for fixing that performance regression
what a great maintainer
fixing The Slab
who uses comfyui for llm
Linux 6.19 Gets Rid Of The Kernel's "Genocide" Function
they probably report because like if 1% linux user care about some project its still lot of people who will stumble upon it in search 

Linux Foundation's Newest Endeavor: The Agentic AI Foundation
phoronix actual news
bad news
AerynOS 2025.12 Brings Many Package Updates
update bring update
linux foundation just companies that use linux
not surprise
ye
linux foundation now have anthropic and openai in it
"linux" foundation 
too many things depend on mcp
already
somehow
isnβt it a year old

mcp who
main character protagonist
model context protocol 
minecraft pigman
fake
neuro sdk but for ai company purposes
meow cmeow pmeow 
most complex pseudointellectual-acronym
MOT conii promise
wrr
mrr crr prr
miyagi coast prefecture
5 days left 
true 
what are you planning 
knife expert
and make it spin faster to shrink the length of the year
can you confirm
year end before christmas
that maximum crab portal
definitely not minimum
minimum number of crabs in portal is 3
maximum crab portal is ocean, silly
ocean surface
can be argued
year end before new year as well 
it will be 2025 forever
sound fun just make sure to start early or they wont update calendar in time
require new year for increment?
no raycasting
expensive
indie dev
anyways time to cry
live in #FF0000
it is not fun

unsmog your frog
bad idea 
<-
<- 
πΈ
i got the ping and was like uhh
rool breakage 
it not edited in push notification 
ye she produce dariacore
that all wanted to tell you


my login manager disappeared when I started up my arch machine again and I was just on a terminal for a half hour googling stuff on my phone
I discovered an update for Ly needed me to go in and manually disable some getty thing and enable ly at that address or whatever
fun stuff all around but now I have my desktop back once more 
Goodmorning 

Cryptographers out there: you have my respect!! This stuff gets wild!
If you like what you see and hear, join the credits! https://patreon.com/zanzlanz
Games featured in demos:
- "Clover" by BugMater: (Requires Flashpoint) flashpoint://run/6d6649bf-2188-43b5-90b3-0ebea635199a
- "Mines" by Tony K Lam: (Requires Flashpoint) flashpoint://run/0ece...

i sure wonder what the p in prng stands for 
well ye i know its not fully random
but it was an interesting video to me on how they revcerse it
n
quantum RNG my beloved
pain 
const ExprTag = enum(u2) {
int,
float,
string,
};
const Expr = struct {
tag: ExprTag,
contents: union {
int: i64,
float: f64,
string: []const u8,
},
fn int(x: i64) @This() { return .{.tag = ExprTag.int, .contents = .{ .int = x } }; }
fn float(x: f64) @This() { return .{.tag = ExprTag.float, .contents = .{ .float = x } }; }
fn string(x: []const u8) @This() { return .{.tag = ExprTag.string, .contents = .{ .string = x } }; }
fn match(this: *const @This(), x: anytype) @TypeOf(x.float(0.0)) {
switch (this.tag) {
ExprTag.int => return x.int(this.contents.int),
ExprTag.float => return x.float(this.contents.float),
ExprTag.string => return x.string(this.contents.string),
}
}
};
pub fn main() !void {
const exprs = [_]Expr { Expr.int(1), Expr.float(2.0), Expr.string("wrrr") };
for (exprs) |expr| expr.match(@as(struct {
i: bool,
fn int(this: *const @This(), x: i64) void { std.debug.print("int {} {}\n", .{this.i, x}); }
fn float(this: *const @This(), x: f64) void { std.debug.print("float {} {}\n", .{this.i, x}); }
fn string(this: *const @This(), x: []const u8) void { std.debug.print("string {} {s}\n", .{this.i, x}); }
}, .{ .i = true }));
}
Nim strongly discourages the use of unsigned integers, as it's considered unnecessary and somewhat unsafe for most applications.

ye pain
not use zig
use anything else
like anything
unsigned integer too easy to underflow competitive programmer not use 
not me 
me battling zig nothing to do with skill issue

for i in 0..<d.num_integers_to_sum
odin
zig opposite of what i like in languages
verbose AND doesnt give much low level control compared to other languages
pointless syntax change
for kicks
Thread_Data :: struct {
what is a ::

stupid symbol many symbol
dont need
sound good
langdev cant decide if .. inclusive or exclusive
unless langdev only know one language

ye that why ..<
,<
thankies 
same
reason me study them
good syntax not exist, good semantics not exist 
not real none of it real
101010101010101
sometimes binaries just 400kb for no reason
rust
me go back to writing words and not programs
healthier for mind
ye dont know what your use case is nothing to add
not know better system programming language than rust and its still not that fun

scala is like rust without tedium
run on jvm
only suggestion but probably wont work for you
sc <- good letter combination

ala <- good letter combination
scala <- mid letter combination
it less than sum of its parts
let me read some day
if what i think it is
deal 

me too disorganized 
you're welcome 


i need to figure out how to mkae a character smile but not make it creepy
nix have that
be nice to each other! be nice to each other! be nice to each o
ye nix do this a lot
i can hear this 
me not seen american spelling except in third party flakes yet
cant think of any specific example though 
ohhh ye
optimise
specialisation 
propaganda by big s
me few both 
Nixpkgs virtualisation
me always misspell 
i wonder why they use uk spelling
leaf ui 
if it was english uk everywhere it would make sense because outside of IT british english is the default but in software american english is

i assume american default for legacy reasons
outside of IT british english is the default
glue ful
i use american spelling because colour sound obnoxious
and z instead of s how pronounce anyway
outside of IT english is evil mix of various dialect and culture
no i don't mean spoken or casual
i use british spelling because
i arrived from mars
and they use it there
colour <- actually martian spelling
mars former british colony 


jupiter also
why former
probably left hand drive too
maybe current
have you seen mars in the last 4 billion years

it very red very sandy
not much going on
sound like london
does there have to be a lot going on for it to be a british colony
- β profitable
- β easy to control
- β make uk look powerful
- β have some important resource
- β have exploitable people


mars not easy to control, unless you consider control as "we sent a lander there one time and it broke down after a few years"

i read ladder

send ladder to mars
does that not count 
uncontested supreme leader of mars
uncontested 100% linux market share
nn me very boring no one knew

were you 
ye waste of bandwidth

im gonna waste your bandwidth
i don't have any more
https://hil-speed.hetzner.com/10GB.bin look at this file






