#programming
1 messages · Page 470 of 1
anything?
anything
bad words?
You know it
I should probably install and set up tony so I can run my commands through it
(sorry swapped tasks for the night 😅 ) so I would juxtapose the two "models" (which I realize was overloaded for me, concept vs LLM; mb
) as:
- the IS **isn't aware of the IL
- the IS is aware of the IL
In both cases the the IL affect the distribution of responses as they are the source of the base weights. The isn't aware case the is closer to unified sense of self (talking to Claude as Claude, then asking Claude for an ephemeral IS: "Hey Claude, what would Dr. Watson do in this situation"). In that situation the LLM is aware that they are an entity that is able to produce simulations of IS's.
Does that answer your question?
can he spell all of this backwards
Do You Want To Know More?
because quiplash has a hecking time limit and I only have so much time to get functions set up
i think this task would kill him
is this a reference
like 60% of anything i ever say is a reference
STARSHIP TROOPERS is NOW PLAYING and can be found to Rent or Buy here: https://bit.ly/2YITUMo
Humans in a fascist, militaristic future wage war with giant alien bugs.
Watch More:
► Subscribe to Now Playing: https://bit.ly/39UENSw
► Need a Fright? Subscribe to Now Scaring: https://bit.ly/39UENSw
► Have less time? Subscribe to Shorts: http...
I might be missing a reference 😅
I'd guess it is an invitation but reference also works
invitation?
I spent so long mistyping the ref was already said
but im already in the banana cult
they were asking if local court jester and recently dead entity named echo would be able to repeat your entire message, backwards
One way to find out
nooooot to pry but what did this mean 


you can talk to echo if you wish but i am going to look at something for a surprisingly long delay probably first
aka finally train the next one cuz i know how this goes if i use the ones that have dishonored their sacred duty by exploding when abused
Attachement: [image.png]
Are we still gambling on tenor links?
Tbf that one is funny when it produces a valid one
ooooohhhhh yayy yes pls

poor mullan
classic link though, chose to pretend to be alphazero so really aiming high
i kinda miss the youtube.com/[hallucination] gamba
I'm honestly really happy with how fast my project is coming along
Let’s play chess
om
i'm still annoyed that what I ended up building was basically openclaw before openclaw
i just didn't take it far enough
if i had kept working on it I probably would've had proper openclaw before openclaw
i was just really limited by the models available to me lol

Whatcha makinnnn
they were so good back then
26
wtf this was way more recent than i thought
what was i looking at
For LLM?
yeh
I have some commands set up, and I just need to set things up so tony can use them
this is so real though
he's only limited by himself, harness is beyond flawless 
Funny enough
I actually already a (mostly) working integration for Survive The Internet
But if what your doing is more general in scope then call me interested
I can opensource my code honestly
FUCKING ANT
first it went on my screen
now its crawling on my trackpad

If you can remember, dm me when you do
Yes
least broad claude research
damn.. 2000 free completions isn't a lot. i switched my local setup to copilot for now to see how good it is compared to qwen locally.
already used 8% just setting it up and testing it (i had 1% from vscode when i was fighting to disable it
)
I didn't train it myself
or perhaps temperature too low so it straight reads the prompt

Lmao, microsoft actually campaign to PAY YOU to use edge as default browser and do bing search 
it's nice knowing how to program...
i got tired of music players giving me ads, or buffering, or switching my playlist, etc.
so i just made my own 
i used bing rewards once and then never again
It's funny that people even say they won't use it even if they get paid 
If I had to guess
I’d guess that you weren’t actually using the channel as context correctly, just the very last message is getting sent to the model
what kind of music player gives you ads?
you were able to get gamepass for free for using bing search everyday a while back
thats not really new for them
They are now giving away mercedes benz 

spotify, pandora, the mp3 players on the app store...
i see... i just use vlc on my phone and strawberry on desktop
vlc on the app store is a little janky but it might just be my phone
well. i have android so no idea how it is on apple hardware
oh yeah apple is just a pain

i made my media player in html/js/css... but apple doesn't like letting you run js
or open html files in safari...
I’m just trying to help
In the early stages of my own bot there were several times where I thought it was handling context correctly but was still acting like this
It doesn’t normally hard fail so the only real way to tell is logging contexts or simple observations

cast the net as wide as possible to maximize the amount of shit it glosses
of course, you need the one random side tangant
i just dont have good models available
they're not right for the tasks I give them
i've tried several
they're just not able to intelligently use what I give them
doesn't matter how hard i prompt
make many mistakes flag was on
You can use Context engineering
Because prompt engineering alone isn't gonna make Ai smart
It doesn't what you want actually, explcity, detailed
That why some repo have like 20+ .md for Ai/llm
Least impressive Claude Mythos demo
claude mythos reasoned romantically that rust minecraft would be a net gain however no methods apparently exist to get it done
the oracle has forseen that gpt 6.9 is the one
Nvidia ceo is cooking with qunatum computing
im not throwing crap at it and prompting it
i did quite a lot
none of the models i have are good
What agentic ai, are you using?
garry tan is cooking github's ingress endpoints
the instruct models cant even do the simplest tasks without glitching out and spamming the same token until they hit their max token length per response
its not agentic
gotta be params
Oh I thought you're using agentic ai
those can be agentic lads
I've never used hugging face before so idk
What model?
i tried a lot
theyre most definitely not
ai agents😭
shadowlilac 😭
it😭
it😭
in the future entertainment will be randomly generated
give me an example of a basic task youd want them to do
I hope, we can upload our mind to robot body in future lmao
Elon musk also wanted to do that
not sure about the mind upload part but full body robot control would be sick
Yeah, it's kinda fun if we can actually do that
pov
But will everyone abusing robot body?
Like I mean some robot body are literal super human
i cant wait till the ripperdoc gives me moutheyes
I wouldn't be surprised...
can i hide a message? this is actively damaging my braincells
opus 4.7
forgor to copy paste the .ps1 blob into the continued training set
isn't this the wrong model for that kind of task anyway? its a simple powershell script thats something gpt5-mini can do.
whenever i want a simple shell script or something i just use the free gpt version 
i use what i have i can't be bothered to open a new window
one would hope that means it'd be trivial for it though
at least it figured it out
i mean sure... but since its so simple it could also cause issues
also it's not really a "simple powershell script" it's a ffmpeg-specific and timing duration is calcuated in powershell
"ok fine i will test instead of just guessing" ive noticed that 4.6 did this
on occasion
would just overconfidently assume it was perfect on a simple thing and never actually run/test
and then it would be like ????
still sounds like a very simple powershell script.
no read the conversation it's stuck on figuring out ffmpeg parameters
that's not even powershell-related
i think were it in cli it wouldve been able to use the help arg
and not be so lost
it shouldve just fetched docs or tried help in its runner
but less inclined to do so in the chat interface
--help would not have helped for vf
i guess...
vf is a whole other matter you'd need to consult the full docs for that
man tho
if only there was a tool the AI could use to do that....
context7 mcp always running on my chest
i still haven't used a single mcp, ever basically.
context7 is the only one i use regularly
the restg if given docs can just be their own mcp server if they really need to lmao, ive watched cc opus after being told there's an api and an mcp available for <thing> will opt to just read find out what endpoints there are and what's documented and then blast out its own api calls sometimes identifying as the mcp server in the requests
this is not true otherwise react native wouldn’t be a thing
or do you mean there isn’t node js for ios, then what’s surprising about that
i tried the same prompt and it appears to just have instantly gotten it without even docs or anything lol
it had a worse time with the ps1 that gneerated the test video
but it got theem both without further turns from me, first prompt exactly the way you asked and then i just said "ok test plz" and dinged done, 103s total O_o
thicc test
github copilot wants to delete my whole function
I did verify it worked but the sound is pure noise so I didn't want to jump scare
"don't need all this extra shit"
main
isnt this what programmers want?
code reduction without affecting how it works

surely that is what copilot is doing here

ai has carefully verified that your program will always segfault
thus it has replaced your main function with *(char*)nullptr = 0xFF
i just like to make sure that it's clean as a whistle
most people dont put in the effort to scrub the grouting all the way down to 0
It knows ur program will segfault either way so it just makes it happen sooner
Let's the segfault be more optimized
my blind ass can't read this
Surely that will segfault 
likely to segfault or sigill
writes to null are UB and can be safely removed by the compiler
you can also try changing it to volatile char, maybe has some impact
alternatively you can just raise(SIGSEGV);

is this normal for a decorator?
"safely"
Yeah I know lol
Fun UB puzzle:
https://gcc.godbolt.org/z/cExT86jeq

isnt it would be better to just set disable no-any in config files? or js linter not have one? (doubt)
-# idk what decorator term means tho
prefer not doing that because I would still like to catch anys
apparently I can't do this tho
according to chatgpt, copilot+claude-sonnet-4.5, and myself (with a lot of slamming my head onto the table involved)
awa
0 votes 
0 minutes 
awa
wait how do you guys add
to poll options?
Neurosynth
award
bot not allow me to recreate poll
Dog poll
ah i see now, it is one emoji only at left side
yea you can't do in the answer text
Toast
bwaa poll
3
6
intel core i7-13700k
3000W ST-Pro Auto StepUp & StepMeow Voltage Transformer Converter, Heavy-Duty AC110/220V Converter with US Standard, Universal, Schuko AC Outlets & DC5VUSB Port by Goldsource[3-YearWarranty]
Also @true hemlock i found out u can buy ~1.2TB ram via intel optane with cpu+motheeboard included for only 1200$
jn atx form factor
bro's about 2 months late 
Is there a cheaper approach
That i missed
For 1tb ram


its just 
I mean ik 1tb ram is possible but not usually in atx form factor
Usually it's eatx or weird huge motherboards

all tr boards are e atx


it's asrock rack still so they mustve built it for oompa loompas
getting optane DCPMM to work is

and

I mean idk ur the expert in this😭
I just saw it's in theory possible
i mean it def is
It has to be possible
Wait what hi iggly



I've spent several months of my life on this
and boi is it worth not being able to spend much on it
I'm trying to set up a compute cluster locally
I found out both of my dgx sparks are 13900k level cpus

But yeah i need 1 ultra high ram box
Even if slow
For hashmap
For mapreduce pipelines
Cause spilling to hdd is slow as shit
wait try benchmark it
Yea sure what should i run
cinebench 2026 prob
Does it run headless
geek dentge 6
geekbench shit at thread scaling tbh
Does cinebench even exist on linux
Afaik it was mac/winslop only
Cause of redshift
yeah lol
nerdge6
anyways where the 13900k performance info come from 
Geekbench

lmao
If u have a better linux benchmark i can run
sysbench? ive never tried tho
7z has a compression benchmark
yeah passmark might work
Can one pirate SPEC2017?
hell yeah
im running this rn
compression is between M4 Pro and Intel Core Ultra 9 285K, while decompression is close to Ryzen 9 9900X3D
running
also fyi running inside docker container cause libncurses doesnt exist anymore in ubuntu 24

banana?
6
9
2
Banana Toast
🍌
what
@ubuntu-user can you confirm this claim
goodmorning
morbing


huh
and with some luck the case too
ok its not so powerful, around i9-11900k/12700H, single threaded around i9-14900. Memory idk why it's so low since it should be 256GB/s
@true hemlock
maybe if i run with host ipc and stuff in docker
its testing single 64 bit channel
oh nvm
its testing all, just the low ones are single threaded fetch
anyways my cpu
is this one of those god tier threadrippers
this bad boi here
yea fair enough
i mean the dgx spark is still good especially that its 100% silent
u cant hear shit even if ur directly at the fan
my aio is pretty silent
makes me question whether the fan is broken lol
does it also test uh
gpu
rip
its still at least roughly in my pcs speed for cpu
which cpu
i7-13700
ah
idk if i can find my 13900 es score
wait @true hemlock it makes me question
does the cpu only have half ram bandwidth
shouldnt it be 256gb/s
on dgx spark
imagine 256gb is allocated onto both cpu and gpu and summed up
nice
damn that's surprisingly far from the 13700 wtf
it shouldn't be much
there's the better RAM latency
8p+8e on 13700 i think?
8p+16e on 13900
2e = 1p in terms of performance
so its basically 12 vs 16 p cores
as long as it works
despite being on engi sample
hm
wait 14900kf is cheap what
so i can pull 270W juice outta the sff cooler
i thought it was like 800$
intel cpus are cheap now
actually
how much?
489 usd converted, or used for 50 bucks less
to aud?
383chf is 683 aud

300chf is basically nothing here
fair
i think ill just save up for another dgx spark lmao
not a good idea though
overhead be shit
unless if these things support like idk
QSFP+
or something
my mom turns off every pc in our house that runs hot or is loud
politically this is the best bet
tune it
to run stuff secretly overnight
tune it better
i cant tune a 2x3090 to run at full load 100% quiet
13900 es at 0.9w
yea but idle
power limit + clock offset
i mean full load
right, completed
lil fun fact 3090 pulls shit tons of power because the gddr6x in total draw somewhat more than the 3090 chip itself
damn wth
hm
then you could try like
doing 3 dgx spark
since they have 2 of those ports
triangle em 
they are meant to be clustered
ill just buy a switch atp
rip another few thousands
hh
don't think switch are worth
when you already have 2 of those ports each
unless if you wanna go beyond 3 of the dgx
obv i'd get 4

i would need to set up 3d torus topology
man
if only i don't have bills to pay
what's good with 15K a month when expenses are almost all of that aswell
twisted torus setup
works with 4 cleanly
gives 40-70% higher increase in bisection bandwidth
@true hemlock i can just expense it through my dads company so he removes it off his taxes lol
lucky ass
bro makes 15k/month what
deepmind researcher
wait 15k aud/month is not much
i was thinking usd
why is global inflation so shit everywhere😂
anyways i wanna 200GbE nic
number keeps increasing every time you mention it 
i think pcie x16 4.0 gets almost fully saturated
yep seems like it
connectx ones are like 1500 no?
nice!
~A$300 for mellanox
New achievement unlocked. All checks have failed
through my distributor
I have a job?
also @true hemlock do u know whether those weird soldered laptop cpus are legit on aliexpress
or do they break
which
modt board?
bro this isn't even an error 
why did it only error now 
I am partial to build script failing as an error
fair enough so they wont just explode/break in 1 day?
i suggest buying the erying one but others might be also legit
this is pnpm install failing
We need to categorize build script as optional and mandatory 
what do you MEANNNNNNNNNNNNNNNNNNNN
pros is that for erying ones they have custom vapor chamber ihs
cons they defaulted to thermal paste for die to ihs, though the board usually comes with liquid metal kit (not the liquid metal itself)
another cons is
its usually B660 chipset
not the high end z series
i mean i just want for cheap nas
yeah just storage with ssd promotion and 10gbe
why do I bother retrying
none of these modt comes with 10GbE usually
i found some nas boards but they are all N100 garbage
they may or may not have 10GbE depends
but there's plenty that also comes with 4x 2.5GbE
nah i need 10gbe
dgx spark has 10gbe
current nas i have is shit 1gbe and hdd only with no ssd cache
and i aint paying 1000$ for a synology upgrade
lmfao
hard to find the one with 10GbE because they're usually SFP+
you can use nic for that though
that takes x4 pcie
wire it with those m.2 to pcie
10G base t sfp+ module
x4 3.0 pcie should able to support 2 or 3 of those 10GbE easily
transcievers my beloved
but yeah i'd very much just recommend Q670/H670 boards, slap a 10GbE nic on it and they usually come with bunch of SFF8643 ports, each could support 4x SATA/SAS devices
wouldn't recommend es/qs on these boards they can get wacky sometimes
probably 12500T for lowest power
since that's what most vendors recommended for such boards
20gbps?
i have 3 empty connectors
most Q670/H670 should have 20Gbps
though idk about 40Gbps it only exists if the specific board has the mux
whats q/h vs just b660/z690
seems the same thing but scuffed china version
Q670/H670 are nas specific chipset
better than B660 and B760
but still a bit worse than z
so more ports and stuff?
god if i could i'd just move out and start stacking 2u servers
also
H670/Q670 uses DMI x8
same as z series
but B series only do DMI x4
(interconnect between cpu and chipset)
ok, can confirmt he ax201 fixed my issues with bluetooth
the ax200 that came stock with the mobo just sucks apparently
Just like with your RAM, the electronic just knows that it will be used by Sam and decided to just not work 
its jsut a known hardware issue with this mobo
they used to RMA the mobos for this, but i don't have the patience nor am i in the warranty period
so i fixed it myself
im glad the wifi and bluetooth thingy is removable
I wish more things were designed with repairability in mind
how do i realistically fit 4070s + 3090 + dual 40GbE nic in my B850 hmmm

glad my B850 can do 8+8 bifurcation
what if instead of fitting more hardware you fit more software in less hardware
So use that 
GPU on M.2 
converting from x4 4.0 to x8 3.0 need custom asic/chipset switch
that it does
honestly the x4 slot can have the Titan v still
4070s + 3090 on 8+8
then i'll figure out the nic with m.2
surely
surely
doing random PCIe nonsense that boards and cases were not at all designed for
i might actually have to get the 90° 16pin connector
i know where to fit 3 gpus
- 1 nic
should be fine with a 100Gb ConnectX-6
seems to have PCIe 4
40GbE nic only uses a x4 no?
PCIe 3.0 x8 
there must be one that has a 4.0 x4
B850 doesnt have a lot of pcie apparently
for a mobo that costs more than my x570 aorus ultra its kinda a let-down
nope
yeah b series are shit port wise
but anyways
its possible
i have like 2 free m.2
what specific mobo is it? i do find some with 3 pcie x16 slots
will do bifurcation on the x16
MSI B850-P
Good ol reballing
happens when its soldering crack that essentially temporarily "resolder" it but it usually will be dead in few weeks again
but 2 of them are x1
just to support weight
ig
the top one is gen 5 at least
i doubt you can bifurcate into 2x gen 4 x16 very easily tho
in terms of bandwidth, ther eis enough available, but gettign it will be hard
you need switch for that
which i ain't doing
i think 8+8 4.0 is enough for both 4070s and 3090
more than enough even
Well, depending on what you use it for no?
If you only use it every once in a while for LLM inference then yeah, not that bad
But if you run render farm 24/7
first slot is gen 5 x16
third slot is gen 4 x4
2nd and 4th slot are gen 3 x1
right?
i should try this... my 1080 also died a few years ago
pcie saturation is still hard in literally any case. x8 4.0 is basically x16 3.0
which 3090 and 4070 super can never even fully saturate
and then
gen 5 x4 to m.2_1
gen 4 x4 to m.2_2
and through the chipset gen 4 x2 to m.2_3
so
yeah
god i didn't even realize the gen 5 m.2
only using 1 m.2?
that reminds me
idk why people are complaining about gen 5 drive running hot and shit
bro
just
slap a m.2 heatsink with active fan on it???
dont most drives have heatsink versions available? and most mobos have built-in heatsinks?
mobo heatsink are thin
i odnt know how to get the temp from my 990pro, but im pretty sure it has been running fine
you want these mf for gen 5 preferably

thats as chonky as my VRM cooler
this one is also thicc and has built in centrifugal fan
i dont think id have space for that without using a gpu riser.
my case also doesnt have a location for io out with a gpu riser
might use this on my main build
cuz why not
you can probably use that gen5 as your main drive and use the other one for gpu or something ye
my drive is gen4
is it not backwards compatible?
honestly i might aswell just put a pcie 5.0 capable gpu on that m.2
it is, and i already put it there
or i guess you can use the gen4 slot and tap of the gen5 m.2 for better gpu bandwidth
ye
i put all my spare m.2 into my pc, except for the one that will go in the new opi5 when i buy it
and the old windows drive in the external enclosure too
||idk why there is a x4 but that keeps speaking to me||
the place that has my mobo is closed till monday 
toast
it
toast
Toast
Toast
I’ve clearly been outgunned
I got excited and thought you meant a that runs in m2 nvme slot
(protection needed)
I don't think guarding against page table modifications will help with this attack
Fun fact: Windows implements this on compatible CPUs
https://techcommunity.microsoft.com/blog/windowsosplatform/protecting-linear-address-translations-with-hypervisor-enforced-paging-translati/4399739
Microsoft is committed to continually raising our security bar
I think there is a typo in here, it should be lowering not raising; don't know how they made that typo
It took until 2026 but Opus 4.7 was finally able to read a pdf instead of skimming keywords and pretending to read it
AGI coming soon™

scientif
your 00:00 is my 01:00 
o
o
oh my god balloon bannan
bloons tower defense 

Why is everyone named Toast?
toast
Toast
toast
toast
Prisma-tic-tac ^.^
not for long
prisma
what
See? Printer are stuff of evil
Actually
I wonder what GCODE Neuro would output 
yes but I'm gonna gatekeep it
chat wtf i just switched my keyboard switches
the keybaord is silent asf now lets actually go
they werent even high end ones, just boguht some off aliexpress for like 30 bucks
The wonders of expiring parent
Patent
Switches used to be patented by Cherry, now that the patent expire, it is free market
bought these
with 20$ clone keycaps off aliexpress as well
its genuinely so good
Key switches are crazy good, for crazy cheap nowadays
my mom finally wont scream at me anymore cause my keyboard is so fucking loud at night
I have a set of boba u4t, boba u4, and epomaker zebra, and it cost me less than £150
Check out my cool watch with central tourbillon
i love mechanical watches
my old apple watch could never

Shadow never beating the rich allegations
bro doesn't have a magnetic kb in 2k26
a feeling like i've said this already
I am the opposite. I daily drive Kailh Jade Box
Here is my personal review and Sound Tests of the Kailh BOX Crystal Jade Clicky Switch. This switch is like the regular Kailh BOX Jade, but on steroids! Louder, more tactile, and heavier. I conducted detailed sound tests with GMK, MT3, and SA keycaps, comparing it with other popular switches in a controlled environment to provide a clear underst...
in theory i do have a magnetic kb
it just hasnt arrived
root drive was a bit more rotund than i expected
smelt like rocm
var/lib/archbuild
1.6TB
rocm build dir is 75% of the drive use :^)
What did you do bro, lmao
lotta buildin

such a small project we have after all
dependency tree
should weight the size of each node by how many lines of code the repository has


Naah, should weight by how burdened the maintainer is 
it will look identical
propose to gource
actually there 0 files with >1k lines
tho there is 2 pure logic files with almost 800 lines 

like
including all rewrites
weight by how many lines changed
some parts of src (left branch) could be bigger than entire right side 
only ~35000 lines net positive in total, excluding data files
-# ~26000 my impact 

Here my prompt
awa
Gemini is dumb so I have to write more prompt to make the game
i know i cant talk cuz i use gemini too, but why would gemini be the issue with the lack of game? 
it outputs only as much as you ask it to
I've fixed some prompt
Gemini doesn't seems to have creativity
Unlike claude code where it suggests like workflow or something I'm missing
hiiii!
But gemini just do it as my .md tells him
do you expect gemini to write the entire game code flawlessly?
No.
ye, thats what it does. gemini does, and only does, what you ask
lets not use ai for codding
Also sorry for my profile being little unhinged
i feel like im talking to 2 people on opposing sides. ai is a nice tool, shouldnt be fully banned. but its just a tool, the programmer still has a job to do
This is me(100%)
welp, your github doxxed yourself again i guess
Also not touching grass
Anonymous github?
nobody forces you to use your real name
dude i found your facebook in 5 seconds
brother. your school and location is on your facebook
do people not care about their safety anymore?
No fear
Oh yeah? And do they gonna pay ticket just to visit bangkok?
https://github.com/SamuelTulach
I mean people put their name/pic into GH profile all the time
you dont need to visit bangkok to abuse your private info
So what?
idk how it is in Thailand, but here, if your home adress is spread online, innocent pranks are people sending pizzas to your house and stuff forcing you to pay for them.
but there have been a lot of cases where houses got raided by the police due to bomb threats and stuff
oh nooooo, poor literal who
i also dont live in the usa

You need to knows how to speak my country language to do that, because english is sus
and it could very easily do that if the programmers took minimal efforts
in this case its not a known person, but im advocating for people to at least try to keep their personal info somewhat secure
interesting integration idea
set up neuro on a fishing boat with motorized poles
FishMCP
Revolutionalizing fishing with agentic AI
Actually I think MCP is out of vogue these days innit
Fuck if I know, I think the Hackernews “”people”” like just giving LLM a CLI and having them figure it out
Most agents just give Garry tan a ring and ask him for stuff and he does it
I can say these info is not always danger.
Some random chinese hacker tryna attack me from nowhere in some big discord server, by letting everyone verify in server to infect thosuands of users including me, hjiacking sessions but I just stopped it and leave the server
well you failed at keeping your personal info safe, I know your real name is ||Banana Toast||
Discord hacking is kinda getting out of hands
there is no point in that when the info is already chived
Yk you can easily get hacked even from trusted big server, by just verify
the point is to reduce the chance of the info being misused
well even if the info’s out there if you scrub it now then it’ll at least take effort to find it
hello shiro
hi shiro

i think the 1st thing you should do when doxxed is not get doxxed
but if that bridge is already burnt

then the 2nd thing you should do is not tell anyone about the dox, ever
3rd thing is to gas light
even if the info is out there, it's only available to the people who know it's out there
so don't go around telling people you got doxxed
true
and if you're a kind individual, don't go around telling people someone else got doxxed
the convo started cuz one of the chatters here shared their github with their facebook and such linked

so its not like they got doxxed by someone else, they're freely sharing it
i see
well if it's freely and intentionally shared
hi I got doxxed once
what happens now that I mentioned it?
now i trawl the internet and find your address so i can send you free merch
the problem is people knowing the info is out there, and if they have a specific agenda against you, they can misuse it
nothing until you piss someone off enough to make them go out of their way to do something
even if you don't have enemies now, you could someday
for most people, their info being public probably wont matter, but you dont know yet if you're one of the people where it will be
especially if your future self will suffer on their birthday
i have a very hard line on what i publicize here
at a point of no return
all I can do is train an AI to kill anyone that threatens me
because my position absolutely invites more attention than is safe
true
@real sierra
awa


awa
actually thought that's in geneva conventions but it got rejected
@real sierra

📣 bweeeeeeeeeeeeeeeeeeeeee
i asked earlier but you prob didnt see. how has being an admin been?
fumoquest host
At this point, everything that isn't yellow and named Toast grabs attention
i can no longer doxx you
not much has changed
Toast
oh ye, whats up with fumoquest?
another orange namefriend vvin
shiro why arent you yellow toast
poorly documented networking library creating poorly understood issues



happy birthday shiro
i cant remember if i said it on your actual birthday
i think i did
no bananas
bad toast
because im not yellow flavoured or toast coloured

you're hungry for attention then I assume?
banana is a colour and a flavour
toast is also a colour and a flavour
and a way of life

banana is a fruit, and i am a banana, which makes me a fruit
as someone who writes about once a week. nah.
you were already


















