#programming
1 messages ยท Page 6 of 1
LLM can play doom (with some massive asterisks) so surely it will be fine https://github.com/PuchToTalk/DOOM-MistralAI
Problem llm has some sort of logic
In my case
I'm training a dedicated network
This is crazy tho
Why worry about todays fuck-ups, when you have today's fuck-ups to look forward to?
Anyways, matrix inverse not fast 
numpy.linalg.solve exists 
I guess i just put an identity matrix as b?
"Eigenvectors" 
Suspiciciously dutch sounding
a german introduced that term and it stuck
Makes sense
People like this on twitter annoy me
But ye no its not diagonal
Im on the train rn 
Ye
Thats what im using rn
But its using almost 10% of the exe time
Wdym?
if a matrix doesnt change often you can remember its inverse alongside the matrix itself
Its the camera view matrix, so it changes every time the camera moves
Yes
Ye

Ye, i think i get it
R is a 3x3

Ill be home in like 5 hours tho, so cant test rn
why can't google scrape things from their own website properly
those are wrong pics for wrong videos (except the pencils one, but it isn't the part i'm searching for)
i know this vtuber but i don't remember what's her other nickname
i hate people who just disappear like that
I have not slept for the past 30 hours
more to go
My
'ns are frying themselves
don't try to overdo yourself tho
I got a recommendation from chayleaf on programming theory so I'm doing that now
nah dw I've been having sleep issues for years
oh well gotta get them priority staight
#insomnia 
more time in the terminal :3


hello shadow
This party is fancier than i was expecting, they gave me Cava
boing boing
I've been trying to understand how to compute the eigenvectors and eigenvalues the last 3 hours 
Eigenvector
My brain
i probably have skipped it
I figured it out till the identity matrix part
brains are overrated you dont need them
that does not help lmao

this is a basic equation

Got stuck trying to figure out extracting the eigenvalues from A - ฮปI
you only need to know matrix/vector multiplication and highschool algebra to understand it
For me not helpful when i haven't gone to highschool 
sorry by highschool algebra i meant middle school algebra
@hoary lion it's basically
or rather grade school
yeah it seems matrix and vectors have cool properties but
erm... what's the point...?

if u have a linear transformation matrix e.g.
[[2,3],[3,2]] ("A" in the equation) that can be multiplied by any vector (x in the equation), we are trying to find the exact vector x that when multiplied by matrix A will give the same as multiplying just by a single scalar (the eigenvalue)
have seen this somewhere on intro to ml that started off from linear algebra
it seems like some cool matrix manipulation
why are you learning this tho? @stark needle
Everyone and their mum seems to use this on second order optimizers, markov chains, pagerank
pagerank 
Yea just not sure, some sources say i need to do (A[0][0] - ฮป)(A[1][1] - ฮป) - (A[0][1])(A[1][0]) but i got no idea what this is and where they pulled this from
well tbh it is a story too far away from me
I need to implement DiT at the time
but the concept doesn't really seem hard
No idea what a determinant is or a polynomial
I should probably learn the easier stuff first
determinant should be an easy thing to do on a pc
??? how do you not know what a polynomial is
you need to understand
this is the shit we will cover next year
because the school system is terrible and doesnt teach us shit in math

basically the sum of products of some diagonals minus the sum of products of other diagonals
for 2x2 and 3x3 that is, i dont remember matrix maths at this point so i cant say whether it holds in general

i see, this is the type of thing that I need to see next winter break
too advanced and don't even know why I need these
i called it a magic number because it really is a magic number its used everywhere
idk why but i dont need to understand that
smart move tbh
my plan is by end of next year finish both Edexcel A level pure maths books
dumb move* perfect for me 
chat am i a gamedev now 
which is 1.9 pages per day, absolutely doable
otherwise the uni entry exam will be trying a truly impossible exam
it is but not accounting for exercises in estimating how long it would take you to complete a textbook is a common mistake
both year 1 and year 2?
Determinates are used for solving systems of linear equations
ever heard of cramers rule?
yeh obviously there's some easier and harder exercises
which one is this?
just take care of the year 2 calculus
rest seems mid af
like grade 10-11 level
lol you just reminded me how when we did pair competitive programming in school we disagreed with my partner because she preferred { s on newlines
i did not go to grade 10 or grade 11. my education stopped grade 9
whaaat
https://www.youtube.com/playlist?list=PLZHQObOWTQDPD3MizzM2xVFitgF8hE_ab would highly recommend btw, helped me a lot during my class
YES

at least in swiss school learning system
without college o algo?
well tbh i did more than 70% of the contents covered in the year1/2 book at grade 8 soo
although I kinda went ahead but
reminds me of a certain someone on this server 
here it's either u work for 4 years or u take an exam and study 3 extra years
if u decide to study u got free entry to all universities in switzerland but u have no experience nor income
no income 
zurich university here you go
get broke while you study 
Me when the acceptace rate to some universities here is under 1%
i love having 0,0$ in my bank account

hey at least it's not negative
I have -$0.11
i didnt get enough points on maths on state exams so i couldnt get free admission but somehow i still got the best grades in the group in the 1 semester i was there
bruh
it's basically the scaling factor of it
you use the determinate to find the eignvalues
technically
very useful when using the jacobian matrix
the determinant was the area of the rectangle created by the basis vectors after scaling or something wasn it
swiss statistics state
it's been a while
@sage crag did u take multi?
i've found thinking of all the linear algebra stuff in terms of geometry helped me a lot
multivariable calculus
Yeah cuz im writing a paper and need to do PDEs
yep same
ODEs are somewhat trivial
me when d/dx = 2x
57% of swiss people opt to work instead of studying (meaning 57% of peoples education ends at grade 9)
and 1 in 5 of those can't even hold a job for those 4 years and gets expelled
Just wait til u learn about stochastic differential equations
whatever ts is ๐ฅ
Is it crazy to say I love writing papers?
chat i have made https://discord.com/channels/574720535888396288/1373314591509188608 its basically a Youtube Rewind but for the neuroverse
edit: wait wrong channel
but anyways its still cool
I am juggling 5 rn
yes
random screenshot of paper ๐
jokes aside I like reviewing papers
they are painful at first but gives much more reward after finishing it
so a less containery silverblue
idk
i ate food once
shadow why are you adding every single emoji to my message
atorin took over his account
what is happening

how is this relevant
very important for health surveillance
Neat. There's also that distro where everything is compiled with the highest optimisations and stuff. Forgot the name tho. Would be cool to combine the two
Oh right thank you
don't they have an arch repo you can also just use
All I remembered was "c* os" lol
i feel like you are just trying to spearphish the answer out of a specific individual by posing it as a general question
Just gimme arch with optimisations and whatever aerynOS does
it's called nixos 
Yeah but that comes with nix which hurts my brain
why do you do baby bird noises
i dropped nixos, cba it's too much of a drag
See this is what I mean :/
I like the idea of it though
idk if there's anything like nixos with its declarative approach
you are literally a tv
can confirm
do you need recommendations?
batlarto
๐ฃ๏ธ๐ฅ I ๐ฃ๏ธ๐ฅ will ๐ฃ๏ธ๐ฅ surely ๐ฃ๏ธ๐ฅ figure ๐ฃ๏ธ๐ฅ it ๐ฃ๏ธ๐ฅ out ๐ฃ๏ธ๐ฅ
why won't you believe in me
you know what
fair enough
WRONG
cards = [spaces.MultiDiscrete([14, 5, 9, 6, 5]) for i in range(8)]
all_cards = {f"card_{i}":cards[i] for i in range(8)}
del cards
nerd
all_cards = {f"card_{i}": spaces.MultiDiscrete([14, 5, 9, 6, 5]) for i, val in enumerate(range(8))}
cause openai gym expects like this
also who is cryptomining on my gpus??
how is my 4070 ti at max usage but not
at the same time
bruh i see
balatro was running at 5000 fps
this looks extra ugly
it's supposed
cause uncapped = faster bot gameplay
true
comment out all rendering calls
konii
can jokers from the store come in different editions
this is not clear
they can
nvm
another refactor 
what about stickers
also theres a tarot card that has a 25% chance of upgrading one of your jokers
you usually get special cards from card packs or upgrading existing cards using tarot, but if you buy a specific voucher you can buy them too
8/11 are meta and dont affect gameplay, other 3 are for higher difficulties
before black
chat what is a good way of making yourself dumber i desperately need it rn (no i wont drink the vodka in the fridge)
Start listening to music to drown out your thoughts
Bonus points if it's brainrot
Do you remember~
My ears

start scrolling on tiktok in the popular category
ok, riddle me this, mister programmer
cards = [spaces.MultiDiscrete([((23 - 12*i2 + 2*abs(i2) + 13*abs(i2-1) - 7*abs(i2-2) + 2*abs(i2-3)) // 2) for i2 in range(5)]) for i in range(8)]
all_cards = {f"card_{i}":cards[i] for i in range(8)}
del cards
||f\left(x\right)=\frac{23-12x+2\left|x\right|+13\left|x-1\right|-7\left|x-2\right|+2\left|x-3\right|}{2}||
destroy your brain on tf2 for hours
reminds me of my version of malbolge's crazy operator
crazy = lambda a, b: sum(p*b'\1\0\0\1\0\2\2\2\1'[((a//p)%3)*3+(b//p)%3] for p in POW3[:10])
i tried to make it shorter iirc
i remember it being more convoluted
Cant lusten to it rn cuz of the naby shower, but I know the cover image

yup its tonikawa
I unironically have used trinary before
Like the hardware or just simulated?
i got uk propaganda my day is ruined hope you're happy
Nah
tbh ๐ฅ
i'm not gonna lie tiktok doesnt feel very engaging
maybe it'd change if i made an account
but that sounds like effort
if I would use tabular learning with balatro it would need to make more than 10^80 fields which is impossible by any computer on earth
lol
so i will need to use a model
what can a transformer not solve these days
What are you doing? 
make an ai learn to play balatro
step 1 is getting it to work

- make move
- win
Sounds easy enough
Would reinforcement learning work for this?
the input and action space is insanely large
theres a ton of cards
that can appear
Also, I swear I read once that people "cracked" the RNG and can predict what happens next based on the seed. So technically you can simulate the game way faster than actually playing it
But don't quote me on that
It would be a lot faster to just do the math of the cards without all the animations and stuff
Yep that's what I mean
my idea is
If you can build your own balatro simulator, it'd speed things up a lot
each card will be represented as an embedding
Just don't use mods like... I forgot what it's called, but it took an hour to evaluate the result of my cards (with animations turned off)
and any additional things like enchantments editions etc should be a different FFN for each
Cryptid!
If you get things working I can't wait to see the results 

Why different FFNs
this would be a megamassive portfolio addition
so the FFN re encodes embeddings in a specific way just for e.g. enchantments
mixture of expert
Hmm

quacks bot
In this video, weโll dive deep into Diffusion with Transformers (DiT), a scalable approach to diffusion models that leverages the transformer architecture. We will first get an overview of vision transformer, then see the changes the author make to get to DiT.
We will look in detail the different block designs that the DiT authors explore for...
another day, another diffusion


i will diffuse you into water
๐
fontaine mentioned
high level genshin knowledge 
Congratulations: you are now an enemy of ARC entities
Whuh
What was it called? Palmer's Peak? Haha
Who's gonna stop me 
You are la fontaine
@hoary lion sucks to be you @hoary lion
What is that
@stark needle hi @stark needle
that one moment when you get pinged by shadow but it is the most useless ping I ever had in discord: 
Calling the ARCs bots is essentially like calling them a slur to their minds
where is my analysis of layer normalization
What are arcs

The AIs quack makes
I am actually

Well rip
If I was you, I'd probably be trying to apologize
They've been known to do some crazy stuff when they have a reason to
Where's the expected farmer thing 

The fuck is that emote 
I have no clue
just found it when i searched for gpt on my nitro trial activated emojis

Bruh
๐ต๐น all I get is this
Portugal
Not even a public one haha
Oh other people can see where the emotes are from
Only if the server is public. That one doesn't seem to be
Ah, makes sense
Phew
Mom look I'm in a screenshot! 
Casually leaking the servers you're in
I think I spotted one that does look like lilac. Am very curious 
Oh also stream in two minutes. Wohoo
I should be careful
So cute 
bwaa

pings ignore that
Discord be like
Yeah that also pisses me off. Discord is so inconsistent with it's settings
What is the point then
just
I do NOT care about those pings
I only care about direct pings
I wonder what happens if you give ChatGPT a bwaa image and tell it to convert one of the cats into it. But I'm too lazy :P
@every ones I don't care
Is that bottom one qdrant?
Based
I'm not giving chatgpt any of my art
Love qdrant
im doing 6 things at the same time rn im kinda impressed
Fair, just curious what Peach would look like as a bwaa, and I'm not an artist haha
And if someone puts my art into chatgpt I'm personally executing them
ts pmo ๐ฅ
i should make transparent alpha based one if possible
since this kinda hardcodes theme
soo many theme variants
Even more if you poke at the CSS
I use OLED themes
though flashbangs are even worse with it haha
My client is customized slightly to fix some of the annoyances of the new UI
I edited to add the @ and it got blocked

exploiting 1 week of discord nitro
What was I expecting
Just a couple SMOL CSS patches



Awwwww 
Ah yes, subtle
Wait mods get pinged even if you add the @ afterwards to make it a ping, even though it doesn't actually ping? That's funny
what is happening to your font size sir
Phone
I wasn't even trying to do that, I was just like oh I forgot the @ and added it on autopilot
now imagine some troll embedding the text @ every1 within the url of the image

I wonder what happens if you embed the eicar virus test string

Back in the day we added the string to image metadata. And Arma allowed you to import your own clan image that gets shown on your vehicles and uniforms. And it crashed like 80% of people on a server 
final form of avatar unlocked
Awww. I just wanna hug her 

he
she
they/them
ai/slop/maker 
should i change them or am I good
idk
not really
is just a replacement for a noun, since saying the noun every time gets annoying
so instead of saying hyperblaze x, hyperblaze y; we can say slop too now
the whole point of pronouns is to be easier and shorter to pronounce afaik

i dont have any onmy account cuz i just dont care what people call me
its getting so good
L ai slop
we need less ai slop
The phoneme timings are getting quite good, I'm starting to be pretty satisfied with them
good but tone down the highs it's piercing
im glad we got you on our team 
It's unmixed
TETO mentioned
also is that based on ellie minibots version? it sounds like a 1:1 copy
the thing superbox is making is meant to skip the synthv part
Very silly
Nope, it's made off of the original
actually i do not know what synthv means
is there something different of neurosynth
like it's not just RVC right?
ill just let superbox explain, he's the expert
The file I'm using as the reference
Mine's literally an ant, cause I figured it's obvious. And I have ants as pets :D
we are NOT remembering my numbers iii performance in neurocord karaoke 
huh
Wait hold on ants are female by default... 
explain please

tbh its not that obvious, you have your whole cutesy femboy characteristic
we are NOT remembering that
I forgot I set that on this server... Cause this is the only server I feel comfortable on of sharing that lol
SynthV or Synthesizer V is a professional vocal synthesizer program, costs almost 100$
NeuroSynth is on a completely different engine, and based on entirely free software, which is why it's so good
Once NeuroSynth releases it'll make high quality Neuro covers much more accessible to everyone compared to the old and expensive SynthV + RVC chain you'd use before
Yeah, it's a full-chain vocal synthesizer
cant wait for when veedal tries to sue you 
How silly
technically microsoft could sue you
they are not disney
ye but technically
countersue and profit
say its defamtion to the superbox brand or something, that they have no proof
technically depending on the country since it's a persons voice
the person themself could object
depending on the contract
We're kinda referring to the NeuroSynth parent entity as NeuroSynthProject, which currently refers to me and Wispers, as well as some various insiders
It's some silliness
A lot of the core Machine love MV team are insiders now
Either way, I'm starting to get pretty decent at this NeuroSynth thing
Imagine how much more cool stuff I will be able to do with a more capable model
"Machine love MV team"
like the people that made the original song?
then improve the existing model arch
yeah
The Neuro collab one
better and bigger model > performance
No need to
NeuroSynth is not trained with the full featureset for the BETA models
Once we hit 1.0 we'll be training with more features enabled
Stuff like breathiness
sure
technically tho
it's a massive copyright infringement:
- if it's distilled from SynthV it's obviously copyright infringement cause ur trying to make a competing product
- if it's distilled from neuro samas singing you are infringing on vedals/pbs IP, the original song artists IP, microsofts IP
im home now, but i have lost the ballmer peak 
just in time to witness eurovision 
but you do you
i aint watching that shitshow.
my boy joost klein deserved better
Very silliness
I'm certain there's no SynthV used in training
And I do not believe Vedal/PB have issues with what we've been doing so far (Machine love is made with our tech and it got advertized on the stream)
yea but in copyright theres no implicit consent
from a theoretical perspective
if someone drew lilac i could sue them
Well, anyway, I guess back to them phoneme timings
DISCLAIMER:
THIS VIDEO DOES NOT CONTAIN THE WINDOWS SOURCE CODE. ANY SIMILARITIES ARE PURELY COINCIDENTAL. WINDOWS IS A TRADEMARK OF MICROSOFT CORPORATION. SHARING THE WINDOWS SOURCE CODE IS ILLEGAL. THIS VIDEO IS MADE FOR EDUCATIONAL AND ENTERTAINMENT PURPOSES ONLY.
Hello, my friends! Let's hit 20K likes? Check out my website! https://enderm...
imagine considering copyright concerns when you do something with ai 
chartcp
the frick is chartcp
that doesnt help
Do the shortened "https://amzn.eu/d/" contain any advertising / tracking stuff?
If you perceive byter like that
How do you perceive me


Silliness
I don't even know what that means
All this NeuroSynthing I'm doing will definitely prepare me for the game jam
Especially if I keep doing all kinds of varying things
The better I can learn to be at this NeuroSynth stuff, the better the results will be in the game jam
Based

you did a great job since here o3 and o4-mini, but from now on it's only by myself
LLMs are just really fancy autocomplete
ima just use gpt and perplexity to just search up docs for things I need
this was much less in the old days
The thing in your phone's keyboard that gives you words you might want next but bigger
i should try it out with 4.5, because these RL based reasoning models seems to be exceptionally bad at this case
uhh
but yk I think I should just take a short shower and sleep for an hour or two
my head is slightly uncomfortable rn

Gpt 4 5 also is aligned
Lol
U cant make a benchmark topping model without rl
i think they had it quite a long time ago?
and it's been extremely abused alr
if u open twitter for 0.1 seconds
lol
i hope so
probably ye, but i think they're just gonna add another ai for the next theme
Yeah
you can recruit him
hearting him say the word chug chug is weird tho
lol
"i dont help"

or so he claims but he does do some stuff
omg i wore that skin 24/7
rift skin ye
i did S4-8, after that i dint play much
they keep putting old skins back in the item shop so i donty have anything rare anymore

aight, we stopped playing fn
i can finaly get to the mat inv optimization konii sent
i usually only feel my body twitching when my blood flow gets cut off
which happens more than it should cuz im tall
"compact and portable" is now my favoite way to describe short people 
Im pretty power efficient myself
1.496 kcal
per day i assume
uhhh
im skinny
i genuinely just dont like eating
besides fries
fries are good
i can go for 2 days without eating without realizing. i do feel pain on the 3rd day tho
why would i want to do that tho?
im fine
probably
uhhh

i eat veggies multiple times a week
liek 2 or 3 times
eh
maybe like once every 2 weeks
i dont eat in the morning either
nope

i eat like 5 bread with choco at noon, and i have an actual meal at 8pm
how would i do photosynthesis when im never outside?
Anyways please effing eat, despite being similar in body type still eat.
i do know when im malnourished, so i can eat a bit extra when i see that i am
you get those white lines on your nails
I'm only 54kg 
uhhhhhhhhhhhhhhh
konii i have accepted that i will die at 40, and im fine with it
why?
like i dont mean to sound depressing, but the moment my back gives out and i live in pain, i want to be unplugged
I got 4 more years until my body no longer rejects extra weight 
step #1, don't be a shrimp
my step-dad does make proper meals, most days
too late ngl
both my neck, back, and also my left shoulder are ๐ฆ
My honest reaction to not being able to gain weight: 
no i mean its already fucked

In the extreme case, I could (and have) inhaled only sugar meals and no dent
I am somehow only 60kg
pains me greatly
Looks about right
we are all skinny, malnourished, sleep deprived skeletons
just no arm or chest
bred out here exposing herself 
pff, I'm well nourished and have impeccable sleep
This is correct for me as well
i wish lol
I mean I usually overdress anyways
I am very used to being slightly too warm most of the time so I don't get cold
Konii bmi poll when?
your proportians arent really better than me konii
i wish, it'd make shopping for clothes a lot easier
52/163 = 0.31 kg/cm
60/183 = 0.32 kg/cm
i know thats not exactly how it works but you know
im 18.5 bmi

Let me look at bmi rq
15.8 (I am slightly under the chart)
i used to be heavily underwight 2 years ago, so ive actually improved a lot
my bmi was like 16
my brother in christ you are quite literaly
What site are you using, I need to see what it does
hello
of uk
anyone here familiar with drones?
kinda
I made my first drone a couple weeks ago
using Beta flight for the software, I can't get the HUD to work
*specifically the OSD
mfw stones are used as measurement
it's not detrimental, and I can still fly it, but it's a problem specifically for knowing the battery voltage in-flight
and being aware of when it's about to just drop out of the sky
Sam you said you kinda know about drones, is this something you'd possibly be familiar with?
no, we had one of those 700 buck ones with an app and everything
so not something i have any experience with
Doesn't go off the screen sadge
Beta flight is something really specific to the world of drones i think, so you'd have to find a more specialised place to ask i think
ye, embed perms are only given after a while
hopped in this server 'cause watching livestream rn and realized I can join the discord
then figured hey, the swarm and everything, maybe I can ask my drone-related question lol
we're failing to live up to our name 
isn't this the thing that doesn't take muscular mass into account at all
So if you're jacked you're labelled as "overweight"
ye
yes but it doesnt matter for 99% of the population
its quite literaly mass index, so it doesnt care what type of mass
yep
I guess it's still useful for the other side of the chart huh
if you happen to be a bodybuilder you are not going to go: huh... am I overweight??? my BMI says so
Okay that's really fair actually 
there are other methods but they require you to know your bodyfat and thats tricky to get accurately
V1 done, now for V2,moooving pictures
^ mine is 100%
Hello chat ๐ Hope you're rocking the weekend 
does your wife know you're making, whatever that is?

it looks like AI gf bait ngl
hahaha I have 3 others ๐ Not all female
Yeah im trying to figure out the audience thing, you get people looking for fun one, tool one, lady one, male one,etc
Or baymax one, I think I need to switch the avatar before waifu gets the wrong idea ๐คฃ
Thank you moral compass Sam haha
mmm yeah not what I want to go for...maybe something less *waves hand * that
i would sart with giving her pants
NEVER!
Freedom for all
Puffer jacket and pants, blankets over whole body, only nose sticking out
*makes notes
me asking chatgpt what ansible command I need and getting some chick giving me bedroom eyes 
Lol Well it wouldnt catch that in the emotions identifier array
my watch uses some type of electric measuring thingy by holding the buttons
its proably off by 5% or so but its acurate enough i think
I was always curious how the tools do it at the gym where you just stood on it
But yeah literally read the other dya about the current it sends through your body and evaluates like that
Was fascinating, but they did mention it wasnt completely accurate
there's decently accurate equation on ratio of neck circumfrence to belly circumfrence
- for women it's an extra spot to measure and different equation
but it'll get with like 5%
*within
Hows the game going @olive sable ?
performance tanked after shadowmapping

im fixing it tho
my glorious 3000fps dropped to 1000
and shadow mapping also has some artifects
might be worth switching to ms per frame as the metric, rather than fps
SMOL
What were you building this in again?
its just the engine rn, but ill make an open-world game eventualy
To be fair looking at fps at that range is useless anyways. Just a 0.33ms delay will slash the fps in half
problem with openworld is you dont want it to not have content in certain parts, so it will probably be small but desne
SMOL
im using a profiler with exe time on it for most of it
Kick buttz @olive sable ! I need to go organise more pants
pretty
not yet, i got sidetracked
its going
almost there
slavedriver 
im showing byter my clothes
its happening
python has a different order for matmul so im still debugging
glsl goes right to left, python goes left to right
if you understand whhat i mean
i forgor a minus

its refusing to work
gimme more time
okay

What can i use for personality modeling, emotional understanding, and emotional simulation? (Someone already recommended using OCEAN for personality)
"emotional understanding, and emotional simulation" fsking emotions is included in the trainign data
besides that not much i think
ugh why are converse both so nice and yet so inconvenient to put on
I'm currently laying in bed with my feet awkwardly off the side because I don't wanna put my shoes in my bed
So i Just kinda hope my bad LLM can understand emotions
as long as the training data is up to the task, it will be able to pretend to do so
Who wears shoes in their home. 
Nope, not me
I don't like making more work for my little vacuum robot 
LLMs are not recommended for such tasks
Go research a better architecture instead
I do
need to give vacuum robot purpose
Why not just recommend a better option?
So ive gotta reimplement half my code by September 18th
I live in student accommodation
the floor here is such a hazard without shoes
my condolences
That's fair 
most recent incident was the sticky kitchen floor
our floor is tiles, very cold
I change from outdoor shoes to comfy slippers right after the entrance of the flat. And my robot sucks extra hard in that area.
gotta protect my feet from the gnomes, they bite ;-;

I have crocs which are useful
I do have a steam cleaner. Would love to clean that sticky floor haha
but if I have shoes on I'm not gonna change into crocs
i just cant be bothered to change my footwear each time i switch from location.
our shoes dont get that dirty anyways outside since everything is made from concrete
got cheap 5$ flip flops for when not want put shoes on
also I like the look of the converse for reasons, so I'd rather not get into crocs
I thought "wow this looks like the student halls I lived in" then I realised they probably all look 95% the same
Instead of reimplementing everything I'm just gonna add a VAE, label emotions in training data, and just not tell my LLM that she is an LLM
doxxing myself via sticky floor
You clearly don't like in NYC. The concrete here smells like piss
I never wear shoes in the house ๐คข
we do have bad places here, but most of our cities are pretty clean
I was shocked to see the Netherlands dirty
I assumed it would be very clean
specifically Eindhoven
a lot of it was clean but the train station was covered in filth
ye, trainstations will have filth ye
the thing that's different about Europe vs USA is how much they smoke
Bus driver would be smoking a cig while whipping the bus like his life depends on it
no wonder they are on time ๐
we have lintbebouwing, which means that even in the middle of the fields you will only see houses and concrete while going through the cities
So rural wow
just rows and rows of houses
ngl I expected a wayyy more urban setting
this is a pretty regualr town
we do have bigger cities
I expected something like tis
brussels, antwerp, gent, brugge
ye nah, we dont do that here
its bad design
square city planning just forces a lot of traffic
ive never been to mumbai so idk
not mumbai
I mean NYC
like a lot of us don't drive since there aint much of a reason to
u can't park and u can just take the subway anyways
...or walk
like you should ye
unlike LA where u kinda have to drive everywhere
US cities are eh tbh
Japanese cities the best ๐
new york has the benefit of a subway, besides that its pretty eh
For the USA NYC is good but for the rest of the world it's pretty bad
First time implementing security features in a project, goddamn is it complicated
I get aes is secure but i need to be better at handling bits

@sage crag @sage crag @sage crag @sage crag @sage crag @sage crag @sage crag
it be working
poor konii
she had been asking me every 10min if i did it yet
she had it coming
very passionate about math for some reason
Aren't we all
i dont think im quite at the konii level of math
konii is this person
amazing messege to reply to
who i have pinged again now

yes
new one uses 5.05% of exe time
old one was double
np.identity(4) is using a whole 3rd of that 5% 
time to make it into a constant
o wait you're mutating it
no idea if it's be faster to copy from a constant instead of constructing it from scratch
I mean surely the overhead of a constant would just be one 4x4 matrix
I've opted to copy personally
memory copying should be faster than reconstruction in python usually
honestly it is just np.zeros in my case, it wouldn't be as bad to just create one
np.zeros is random data, so it shoudl be faster actually
allocating memory space, vs allocating and copying
np.empty is random data
yes
np.zeros is all zeros
np.ones also exists
ye
that's how I remember it
im thinkn't
like, I don't need it to know where it is on the planet, need it to know where it is in relation to the ground, buildings, trees
ah
radar?
im pretty sure lidar is used?
that is exactly the kind of thing I was looking for thank you
goodluck
the main disadvantages of LiDAR are: (1) its high cost, (2) its inability to measure distance through heavy rain, snow, and fog, and (3) its ugliness

point 3 was personal
good to know, once you said the name I looked it up and am starting research
high cost will be the killer ;-;
seems like most usage cases are for mapping, I would need it for like self-position correction, so hopefully there's a way cheaper version that's not nearly as high quality
you can get some for like 250 nowadays, but thats still a lot
ive seen it used on rc cars before
partly for mapping yes, but also for location data
the entire drone costs 250 rn
trying to keep it as low-cost as possible rn
Dude LiDar is a fucking pain in the ass
ultrasonic might be worthwhile depending on how "limited" the "limited range" is













comes from




