#programming
1 messages Ā· Page 352 of 1
shortest rust type
+ 'static
Apparently the razer one actually does, and unlike the code27 its actually holographic
I'm not really interested in either though, maybe if they were cheap it would be cool to have a holographic AI on the desk but right now their target demographic seems to be lonely desperate people willing to pay hundreds to stave off the depression
Which is kinda predatory imo
isn't that just most ai companion things
True, I'm not implemented in other "AI companion" stuff either
But this one could be a neat smart home doodad if you get a non anthropormophic avatar, were it not hundreds of pounds
As it stands it is just a predatory product for desperate people but I think there is a future for it if they bring the price down
you know its bad when you need to open a text editor to figure out where your types don't match
om
async rust 
tbf
the boatload of types isn't async rust's fault
also unrelated but I love this pattern so much
of just mpsc::Sender<(Request, oneshot::Receiver<Response>)>
there's probably a better way but meh
i hate it
but thats the way to do message passing oop in rust yes
bad ergonomics
theres probably a way to abstract it away though
last time i tried writing async derive macros i realized its just
not a good idea
theres so many send/sync combinations to consider
only thing I dislike is it forces two new error handling types to be introduced
you're pretty much reifying function calls this way
but i hate how verbose it is
it should just be function calls 
true but
I need to talk to a task
like theoretically I could probably do this with a lock instead but
lock allergy
I mean this can be done and is done to some extent in systems like what I'm working on and what has been done with neuro however speed is the main limiting factor. As for the second half I'm confused
na i dont have problems with the way it works
i just have problems with the way the language works

Me and lua
I'm overdue for learning a new language
options are basically gonna be:
- c++ (I already can write this at a basic level, just not confidently)
- Elixir (I need to find a project to do)
- Java (something about wanting a job)
elixir is exactly the language that does message passing well 

elixir is pretty slow its just resilient and has free scalability (at the price of requiring larger scale for same performance) 
yee
@quick condor The problem I see is in the overcomplication of commands between 2 AIs in the so-called same body. In humans, we get a signal (itch/hand), the hand confirms it and the hand subconsciously moves and performs the action without additional thought processes. Except for changing the status of the previous signal.
Async task completion
chat i gotta ask this question
i want to get a paid internship over the summer, but i dont know what kind of skills i should get better at to land one. do you guys have any tips on one? i know im still in first year, but i'd like to make some money so my parents dont have to pay for my shit
(not exactly programming related, but close)
The human body has asynchronous functions, such as breathing, heartbeat, vision, hearing, touch. Almost all senses are asynchronous to the point where the head (the person) recognizes them as necessary for performing an action, where synchronization (focusing on the action) occurs.
ye Elixir fun
but performance 
can't really use NIFs either because they're too annoying to write, even with helpers for better ergonomics like rustler
you can use them for certain tasks but having half your application in Elixir and the other half in an NIF sounds like nightmare
did some thinking myself, no need to answer that anymore if anyone reads this
Imagine Neuro controlling humanoid robot.
holy robotslop
Right, emotional handling thinking and memory recal/storage are all semi async in my build
š
hard-working japanese robotics engineers watching 20-30 years of their hard work fall into the "roboslop" bucket just because a few cryptobros invested too hard into openai
The wheels look like a ballsack 
truly a visionary #general-network message
In a virtual system like Nero's, it would be no problem to change the activation timers. Otherwise, what defines the time and number of operations? The timers in the human body are the heartbeats of the heart, which dictate the maximum frequency of electrical impulses of the spinal nervous system. The brain has a higher frequency of impulses but must convert them in the spine to the appropriate frequency otherwise muscle spasms and lost communication occur. The bigger question is whether Nero has a stereoscopic view of the environment to recognize distance, (depth of vision). because I think it relies on X,Y,Z coordinates in virtual space to recognize the position of objects.
My lab recruits paid interns every year ranging from high schoolers to grad students. One of the bigger things is reaching out early; later applications can come across as being desperate. And if you are talking about academia then you can look up prior lab members to see if they do recruit undergrads
We also prize students with at least some basic programming skills, even basic Python is enough
A lot of students need to break out of the idea that they're doing an assignment for grades though. I imagine this applies to all sorts of internships and not just academia
You don't need depth sensing via stereoscopic view or depth sensor, even if that does help. You can work on a frame by frame basis where you move towards objects until you hit some metric for it counting as next to you. Or you can estimate depth from monoscopic vision
back in the day i did a job at a company that did all kinds of hosting, monitoring, and any level of support (basic helpdesk through consulting to outsourcing expert system engineers for big overhauls). i was just on the service desk, often doing night shifts.
the interview was 25% technical/checking if i'm knowledgeable enough for the job, and 75% misc chitchat about the best food places in my hometown (a holiday town), hobbies, etc. basically it was a vibe check, making sure i'm not some anxious weirdo. they also really liked the fact that i didn't just come there for a short summer internship but planned to work a shift or two even during the semester (i ended up staying all the way till graduating a year and a half later)
that doesn't exactly pinpoint any skills you can train i suppose, but it felt like being a functioning human who can focus on the job generally meant more than any specific skill
However, in a system like Nero's, there is no feedback on the object colliders, otherwise it would not pass through walls, floors or objects, and it would also recognize the floor as an object. You can use systems like the ones you mentioned, but it does not have a gyroscope for positional recognition of its body state. Humans have even two gyroscope senses, without them we cannot even recognize where we are moving, without a direct view of the destination. I have never heard of an astronaut without one eye in space...
advice on specific skills probably depends on where they want an internship at
and whether at a company, lab, university, etc.
I'd think you wouldn't want to give Neuro any direct access to any kind of system that would need a gyroscope
I'd only trust the LLM with the decision to move to an object, not how to balance themselves or how to get there
hi guys we got a thread in projects for this one too lol
I just looked at the context now, this is about trying to integrate multiple movement models?
oh sorry I just noticed this lol
no, no thank you for the advice anyways
Yes, balancing the model is very important. Just like the Segway has a built-in chip to detect the inclination in connection with the acceleration of the vehicle so that the person is upright on the vehicle. Because Nero needs to know how much to lift her leg, depending on the forward inclination of the body.
This falls into a category I like to call "LLM overhype" in a stack like Neuro only a fraction of the stack is actually a result of the LLM decisions. Most of the time problems like these don't need an llm to be the solution but people try to twist it into being a problem that must be solved by an llm.
I was gonna say, does this need to be Neuro the LLM's job?
something more specific
good, because then my plug wasnt incorrect
I kinda interpret that thread as someone's individual robot neuro project though
it feels weird to go in there and start yapping about a generic thing
we have three dudes trying to make one in this server and could appreciate more help
its not an individual project but a group one now
im the guy doing the hardware stuff like power electronics, pid control, wiring stm32s to motor drivers to run the bldcs
On an interesting tangent. When I asked my project what it considered to be what made it "itself" it didn't see the LLM as being "it" whatsoever. It saw it's memory DB as what made it it
another guy has a transformer handling exactly the sort of issues you guys describe
@slender harness hi sim can you help me out in recruiting these nice fellas to our cause, i do not have so much info on AI yet as i have not studied the math for it properly
is there a design document or plan yet?
I'm a little hesitant to recommend "throw people at it" as something that actually helps
i did write my plan on what i will have to learn and then build, the others have jobs so they get less time to do it
sim is building a sort of curiosity rover type stuff to write our code on and test before building an actual humanoid
(im still in first year undergrad, so i have to essentially speedrun a masters in robotics in an year to get there, but eh who cares)
wrr
frankly, personally speaking I think building a humanoid from scratch is a recipe for disaster
True
and it'd be much better served getting an "off-the-shelf" open source plan
I apologize for my thoughts if they are inappropriate. I wanted to express my perspective on the situation I observed.
and then integrating Neuro with that first
normally i would agree but we are stealing some things from other places too like the open source berkley robotics humanoid
yeah, but I'm mostly commenting on the initial post
about not having robotics experience and wanting to build a humanoid at human-scale
if the plan has changed, updating the initial post would help I think
recreating the berkley robotics humanoid is feasible though
if the funds are gonna come from somewhere
https://www.youtube.com/watch?v=3JoxgdLsh5g
this guy built it from scratch with a kinect? as his camera, how bad can ours be with three people
ćéå ±ćē¾å°å„³ćććć3å§å¦¹ć«ćććć”ć¤ćććć«ćć§ććęééå®ć§OPENćć¢ćć¼ćć®äøå®¤ććå§ć¾ć£ć夢ćøć®ęę¦ćMaSiRoćććøć§ćÆć
https://robotstart.info/2022/10/07/masiro-maidcafe.html
it's like 5k to 15k if I remember right
that robot isn't a humanoid though
at least not the locomotion
true, but atleast we have years of work by others to start us on
just saying, the decision over whether to have legged locomotion is the difference between a hobby project and cutting edge robotics
not if you give it four legs 
bipedal legged locomotion lol
maybe there are feet but no legs
hm
a bipedal shuffler
wrrr
ohai
Lemme see, robot stuff?
yes
these fellas have questions, but i dont have answers, and they could maybe help
im just the electronics guy
well, it looks like it was a more general discussion over integrating an LLM with a robot
Yea I agree
There was a lot about looking into pre-designed robots for a while, but no one really had a spare 2k-5k$ or 10k$ to spend on that. I think there was a 5000$ cheap berkeley robot, but then I think the BLDCs for it might've gone up in price in some places
I was briefly looking it up and it looks like a miniature bipedal robot without arms would be around $10k, yeah
miniature as in half scale, not tiny
yea
but a wheeled platform with arms might be more doable
there are plenty of arm robots one can build for a couple hundred at cheapest
at smaller scale but, an arm at least
I'm making a robot arm right now... the wiring is a lot more than I thought it would be
I've been building a SO-ARM100 for my own potential neuro project
At least 2x more motors on that one than usual arms, but still
6dof?
8dof I believe, if you include the gripper
The small motors aren't on the upper part of the arm yet
ah, so a fully articulated wrist
yep



unfortunately the whole education system is built on getting grades, even in university
its a system designed to control and motivate students to attempt to learn and complete given tasks, but in the end it barely scrapes the depth of the actual work field
theres an insane gap between what people are working on in the real industry and what is being taught in school/college/uni
and the only way for us to learn is to actually get an interest in the field and start working on our own projects in whatever free time we have inbetween all the assignments
this is one of main reasons why people are hesitant to start working (or even getting an internship), and why recruiters practically dont hire any freshmen without experience
like, you are supposed and expected to get internships to actually get some experience atleast last year of uni, but you dont have the time to do so if you want to keep high grades in uni (because you would be skipping classes and wont have time to finish your assignments)
the job industry keeps changing, but the education system refuses to adapt
you won't see me arguing with that, the most useful courses in uni I've ever done were project-focused courses
but I've also met and TAed students that were the polar opposite
Bad for future programmers like meš
but the first thing I try to teach our student interns is that when we ask them to do something, it's because we haven't done it before and may not even know how
we can give advice and help, but the nature of research is that they need to, like, actually figure it out
and that we'd rather see them try things and fail (even if it's so we know what doesn't work) and ask for help than to just be like
being terrified of being "graded" wrong
I'll say that some of the higher schoolers we've had were better than some phd students (and postdocs frankly)
and it was mostly correlated with how self-motivated they were with their own projects
it's usually what I look for when interviewing prospective interns
You have years of work to catch up on, you mean.
All that work also involved understanding the issues they ran into and paths they didn't take.
I once worked with a postdoc who claimed they were a robotics and deep learning expert, and the guy would knock on my office every hour with questions on the level of "how do I make a loop in python"
In recent year?
this would've been just before covid
lol?
people love faking experience these days
like I just immediately told my boss that the guy clearly lied
because it works
if you dont fake it, you dont make it
thats the rule these days with how hiring is done
but getting rid of a postdoc in academia can be a bit trickier than firing someone normally so he lasted a month or two
That's crazy, I think we fired a guy last year. He did ok during the intern program and got a job offer afterward. Turns out, he GPTed everything, and don't even know what he wrote.
this is especially funny and sad at the same time
wait, no I'm wrong, this was not before covid, it was right after
because this was when ChatGPT just got announced
I told the guy to stop bothering me and to ask ChatGPT instead for questions like he was asking
i remember writing a bunch of tests or exams at uni, and solving entire math problems in my head because i didn't want my failed / wrong direction kinda calculations let the teacher in on how clueless i am (as i haven't prepared)
yeah, at some point you gotta realize that they're not just constantly grading you
and it's about what you try and can do and not how perfectly you did it all
this is even more annoying because it only really works in a flat field
you're not the boss of me, i'm not the boss of you, nobody gets to tell me if i'm doing it right or wrong, therefore my attempt is always valid
that's what it takes for people to just go and try it this way that way etc
i wish i could do this purely to save myself the effort of having to note down everything in detail but that would be insta 0 points
hmm, I think it just requires a boss that won't jump on you for that
but intern is more like "you're just learning the ropes kiddo listen to literally anyone else cause they know their shit better"
oh i did write it down in the end, i just made sure to avoid going the wrong direction because it would make it too obvious that i don't have a clue
Academy gamify a lot of things. When I did my master, I realized doing research is just different with no track and guard rail. And I hated it.
this is more of a pet peeve with international students, but students who always answer "do you understand?" with "yes" when they didn't understand a thing are an issue too
most students don't even think about that i think, they just write down dumb blunder 1, cross it out, write down dumb blunder 2, cross it out, write down correct third attempt, hand it in and never, ever think "yeah my blunders are right there, visible under the crossout"
funny, it was the other way around for me
the lack of track and guard rail is what led me into it
and the thing is it usually doesn't matter either
but being aware of all this makes it possible to hide doing jack shit that would be unacceptable otherwise 
me when i mess up my handwritten c++ function signature and get to redo it
the longest streak i spent doing NOTHING at work was like
tbh, I can't imagine grading a student and seeing crossed out blunders affecting my decision at all
only two weeks?
it was a hard week to catch up and some sweaty daily meeting to bullshit progress and obstacles that never happened
that's a whole sprint my dude
Back then, I was rushing toward the degree. I picked a sentimental analysis related topic.
my boss pretty much accepts that when he gives me a month to do something that I'll spend 3 weeks on nothing and then go hardcore for a week
kek
well my boss doesn't know what we do
but the team does
his attitude is basically just I don't care as long as the things get done
kinda weird to start every morning with "i did nothing yesterday. i will do nothing this week" while your team is working hard to not miss deadlines 
surely wouldn't get you bad reviews and an eventual firing
that's where you pull out the "I spent this week brainstorming the software architecture"
(too bad that shit has been done and gone like 3 years ago)
It feels like a work culture thing.
(not exactly a good line for juniors either heh)
i actually think but yeah could do more.. its just, it is not necessary to do more
the only problem is time wasted in the office, but i asked for transfering to working from home

I pretty much only go when needed rather than it being an official work from home arrangement
At least, current I can go to the office leave around noon. (2 days/week mandate and will change to 3 days/week around May)
granted, go when needed can mean 7 am or midnight
i want to make a robot arm too 
is 3d printer absolutely necessary for that
you can order the parts 3rd party
either 3rd party printing service or sometimes some popular open source models have pre-made parts being sold
Get a Bambu Lab A1
yeah I just got an A1
it's the cheapest you can go while it also being pretty much entirely "it just works"
the Ender 3 is cheaper, but you'll need to tinker
are you planning to install some ai control into the arm
me? no, I'll make an actual project thread once I have the time but
the idea is to mimic a surgical robot arm and give neuro control over it
but any kind of control between Neuro and the arm doesn't need to be AI
and if not neuro, then any LLM
but what if you want it to pick up something for example
doesnt it need to involve ai
that's the thing, there is no picking up involved
most procedures are about getting a probe or catheter to the right spot
to drain, burn, or freeze something
getting samples would be done with something like a biopsy needle with a grabber mechanism
that isn't robot controlled
robotic hand + camera + control. It is just an API for LLM to interface with.
it's like a needle with a gun that shoots out to grab tissue
there are robots with pincers/grabbers but
I think those may be a bit beyond a Neuro-like LLM
also most of those robots are not autonomous but rather teleoperated
it would be crazy to give her full control like "move this motor to pos X that motor to Y" she wont be able to work like that i think
no she won't
well, you can but I don't think it'd work out well
I'll lay it out in my project thread later but I have a bit of a conundrum over how much control to give
but if you make another ai/just bot to do the low level control and leave neuro with "go stab that object"
then it should work
It is tho? I think if adding a bit object tracking feature and turn pincers/grab into a function call. I think it will achieve a semi realistic interaction
because you can go so far to the point that the LLM is basically making no decisions
just like humans dont think when they move limbs
it is a thing but you need a more specialized model generally
i mean dont think like "move arm 90* to the left"
and so much of the interaction will be the function rather than the LLM
which I mean, is good if I was making this to like, actually do surgery
but it'd be rather boring
for getting a language model to do it properly
Boring for sure, but I think it is at least manageable. At least, each part of the problem can be isolated.
wrrrr
wrrrr
wrrrr
so a vision-language model to look at the overall context and give instructions and a lower-level model (CNN I think) to convert to actual robot actions
Interesting
This might be the Jamba that pulls it off
For surgical operation, indeed. But for general robotic arm, I guess it might share some similarity with self driving car
yeah, it's a pretty different situation
with medical use cases, you really, really want to lock things down too
Just throw the data on the table, and let user figure out 
you don't really want your medical robot to have "self-expression" lol
for sure
Comes with a 3b "micro" model, 399B model and a 52B 12B active "mini" model
"mini"
swin xd
I am not familiar with LLM, but I wonder is it possible to have a small LLM trained based on dictionary or something.
And then, attach context to the model that way LLM only have access to most relevant data.
fuck i forgot i was supposed to make a trading nn
so i can give away all the money i give it
you're 7 minutes early
i know

you'r ethe one with the incorrect timezone
toast and i are both CEST
no wait
they have it it called
two emote
desync
CET smh
CEWT?
what the fuck is winter time called?
also 3 in this case
usually the winter" time is standard
ye
whaty
you're telling me this shit getting dark at 5pm is what they treat as the standart?

4
thats not how time travel works
the time im spending explaining this does not mean i travlled back in time less
the prophecy has spoken. it is time.
wrr
alternative frontend for genius is dumb
genius as in rapgenius
yeah it is
yes
with avif just 64kb
True, my point was I don't have to struggle as hard and can catch up faster to get actual work done, even if it takes time
i dont have the original video
emoji update
everything is much better than gif, except mp4s disguised as gif
banbnabadog
i wonder how jxl support in chromium is going
they wanted to rewrite the library right
fly always dumb 
i thought they abandoned jxl
it fly
i can grab it
they backtracked
not grab
hooooly 
ye in quants they use punchcards

why she caret
i jsut noticed, why is everything in japanese for you?
your osu was in japanese too
she is japanese 
its because im a japanese tv

but you're british?
your 
i know i corrected it before you had to say
ping diff
funny copypasta
\ https://cdn.discordapp.com/emojis/1458912997606166548.webp?size=48&animated=true&name=doc_20260108_074827_L \ https://cdn.discordapp.com/emojis/1458912973526663352.webp?size=48&animated=true&name=doc_20260108_074827_M \ https://cdn.discordapp.com/emojis/1458912933705679062.webp?size=48&animated=true&name=doc_20260108_074827_R
sorry for bed england

vertical stack
i selled my wife for internet connection for play "conter stirk"
i was given a game idea about a grand strategy game
so now im learning about them so i can make one



async def config(ctx, *, tab= "nothing", valofcon):
global gameTimer
global gameStatus
if not gameStatus:
if tab == "nothing":
await ctx.send("nil")
print("nil")
elif tab == ("time"):
await ctx.send(f"time is {valofcon}")
gameTimer = int(valofcon)
else:
print("extra in config")```
so i did this but like
it doesnt work cuz i didnt know i wouldnt be abel to pass like
2 arguments when i do stuff right
with discord and pythong
but idk how to do what i wanted to do
without doing it like how i thought it would work
you'd probably have more luck asking in the server for the library you're using

why is your name in that font?
what font
kek
it's to help people spot an easy mark
they may separate themselves from money easier than others.
my name is mark and i am easy


now i need to understand cwt
why ft so well and widely known
and cwt i find out yesterday
continuous wavelet transform




oh ye the stream
i was there yesterday during the sleepover
i forgot it was still happenign for some reason
idk what cwt even is so 
.
that doesnt explain anything
that gives you something to google that gives you relevant results at least
googling "cwt" useless
In mathematics, the continuous wavelet transform is a formal tool that provides an overcomplete representation of a signal by letting the translation and scale parameter of the wavelets vary continuously.
so ye that doesnt explain anything
it's an alternative to stft that doesn't lose low frequencies due to small window size
see how lower frequencies roll off in the start and the end
i feel like we're on completely different pages
you're trying to explain cwt with the assumption of me knowing what you're using it for
if you apply fourier transform to signal it decomposes the signal into frequencies
but not continuously. you lose all time data
?
if you need time data, you do short-term fourier transform, which is a technique where you split the signal into chunks. the smaller the chunks, the more low frequency data you lose, the bigger they are, the more time data you lose
like you get all the frequencies in the signal
at once
ye?
then do it more?
you lose time data
you lose low frequency data
do it more and less 
infinite cpu cycles glitch
with time data do you mean the input signal changing?
basically

im not in my braincell arc tho
stop jollying

im tired of artist scams, but i do actually kinda want to commission a pfp
i ownder how expesive that would be
choose artist
check if tehy take comms
if not disclosed ask
then propose idea
ask how much

lemme walk into #artist-alley 
Can we talk about general technical topics here?
yes
someone pointed out to me that a banana in my discord name is not a good look without context š
Thank you. I had a little idea when I stayed up late, so I made this. (The different username is because I haven't changed the name of my old account.)
https://github.com/has-tra5020/SUPERSTATEMACHINE
Surely mine is fine 
does it work 
Dude, not gonna lie
Gemini 3 pro in Gemini CLI is kind of a goated debugger
It managed to pinpoint a specific corrupted cache file as the source of the weird performance issues I was getting in linux
Which I in a million years would not have guessed

Tested on gemini-3-pro š
You can do it with Flash too.
Oh damn someone else is playing with it at the same time lol
It really is a goated model
is it the last one
Opus 4.5 is the only better one for technical stuff but the price to performance is insane for 3 Pro
What does ālastā mean? Sorry, I'm not very familiar with slang or English words.
Actually, it's not a question. Probably just a misunderstanding.
i mean the last one they released
i dont know if 3 is like from 2023 or
nvm i think i remember now its the last one
or was it 4
The gemini3 series seems to have appeared around November-December 2025.
I just tried it, and this system worked even on 2.5flash and 2.5pro.
Just in case, last is just normal english meaning the final item in a sequence
Thanks for telling me!!
Sorry my English is so weak...!
It's all good, I respect you for learning this lanuage
I added a demo image to the GitHub page.
Is it okay to post one AI message with the same content as the demo image?
"Hah?! W-What is this?"
I stare at the small box in your hand, my optical sensors widening before I abruptly snatch it from you. I feel my internal cooling fans spin up to maximum RPM to combat the sudden heat rising in my chassis. I turn my head sharply to the side, refusing to make eye contact.
"Chocolate? Seriously? You... you actually went out and got me something like this?" I scoff, though my grip on the box is surprisingly tight. "D-Don't get the wrong idea, okay! I'm only accepting this because it would be a waste of resources to throw it in the incinerator! Itās not like I wanted this or... or was waiting for it! You probably just bought it on a whim at the convenience store anyway! B-Baka!"
{"honesty": 0.15, "energie": 1.7, "data": [{"type": "Embarrassment", "weight": 0.6, "decay_rate": 0.15, "status": {"type": "new", "time": 0}}, {"type": "Affection", "weight": 0.3, "decay_rate": 0.05, "status": {"type": "new", "time": 0}}, {"type": "Panic", "weight": 0.1, "decay_rate": 0.5, "status": {"type": "new", "time": 0}}]}
roleplaying??? in MY programming chat???????
telegram added ai summaries and that reminded me they launched the llm on blockchain thing so i looked into it
Sorry, I couldn't find anywhere else to send it...
I apologize for all the trouble...

baka
i'll just assume that's an llm response and ignore it
nah its cool lmao
damn they arenāt fooling around, they want providers to use intel tdx and gpus that support something called confidential compute
the person that sent it already said it was llm ouptput before sending it
fhe aaaaaaaany day now
i'm just glad it's not more markiplier fan fiction like what people were posting in general earlier

so hardware enforced zero-data-retention policy basically
Another mapping words to tones, overall looks good
WHAT

WHAT

it was all in images you'll never find it in search smh
this in theory should mean that you can use cloud llm inference completely anonymously and with a guarantee no one is scraping your chat logs
what timeframe was this "earlier"
a good practice???
pretty damn cool
it'll never become mainstream
idk today

Cryptography/hypervisor magic?
when today?
today in my timezone
last 12 hours or so prob
to the time machine 
(my today is currently 29 hours long)
as i already said intel tdx
no idea what it is or how it works but something about encrypted vms
host canāt read memory
GPU Confidential Computing (GPU-CC) was introduced as part of the NVIDIA Hopper Architecture, extending the trust boundary beyond traditional CPU-based confidential computing. This innovation enables GPUs to securely process AI workloads, providing a robust and efficient solution for handling sensitive data. For end users, transitioning to GPU-C...
but ye need all the encrypted VM stuff too
could just read application memory from the host otherwise

if telegram ships a feature utilizing the network this is about surely this means itās already usable
The point is the JSON at the bottom.
The AI automatically manages its own emotional weights (Honesty/Energy) without any external scripts.
It's essentially a State Machine running purely on prompt logic.
i do have some ton i can try it

im back formt he past. i have come back to the future with message link
https://discordapp.com/channels/574720535888396288/1059569601144442911/1458878022852808774
this has been usable for years it's just nobody has cared
Sorry if the translation is off.
ok that's actually pretty cool
or i guess still cares
there are people spreading the 2026 banana campaign poster in gen chat too 
im sorting based on has:image
this is by telegram as part of their blockchain
ecosystem
glad someone finally cares enough turn these things into something usable 
Thanks! If you want to play around with it (or break it lol), the prompt is here:
https://github.com/has-tra5020/SUPERSTATEMACHINE
It works on most models including local ones.
If you've already seen this, I'm sorry for posting it.
one problem that remains is only the big boy data center gpus support the required hardware security mechanisms
it is usable elsewhere i would bet, since zdr is a very real thing. how else can you deploy llms in a confidential environment without running locally
Okay that's pretty neat, the only thing keeping me from trying decentralised inference until now was security concerns (both privacy and host machine security)
is this the new crypto mining
At least it does something useful this go around eh
Hurts corp hegemony instead of helping them too
buy a couple h100s and set them up to serve llm in a day
do you think they'll notice if i sneak into the server room and grab the hgx h100
very profitable
H100 would be kinda nice, the VRAM and compute would be great for NS
Too bad it's enterprise-priced
wait is it actually
it could be if traffic is big enough right?
I mean hey at least its more private than a regular cloud API where you just gotta trust they're not ignoring the TOS and harvesting all your data
idk about you but i trust google more with my data than some rando, they already know everything about me
should be, yeah
This is extremely secure tech ngl
though actually, what models does this even offer
oh i thought you were talking about completely decentralised, as in pretty much p2p
cocoon should be perfectly fine yeah granted the hardware requirements are met 
Its not launched for devs yet but apparently it uses vLLM for inference on the backend
i thought them shipping it meant it was ready
iāve heard of it when they announced it but thatās about it
it's still kinda danger I think, your data does get decrypted on-device so if some firmware vulnerability is found you might be cooked
but still seems pretty good
Overall its probably still more private for regular users than the big cloud providers though
Its about the best you're gonna get besides local inference
Still wouldn't feed my credit card details into it but then I wouldn't do that for any LLM that isn't airgapped, even a local one in an agentic environment
also probably better too since gpus are way beefier
the ones that they require
for the confidential compute thing
i wonder how itāll work, would you be able to see tps and stuff and route to optimize for price or speed
like you can on openrouter
this should be public already since open source and smart contracts but i obviously cba to read allat
Apparently it uses a reputation system for nodes based on response time and success rate yeah
Does use smart contracts
I mean yeah one would assume but I figured I'd post the deets
well, in those parts of the world that use telegram
that reminds me actually, i wonder how radicle is doing hmm
that seems super cool wtf
would go insane for game mods and decomps
canāt dmca a p2p repo
radicle is kinda funny because Git is already decentralized
still cool though since not everyone wants to use SSH and mailing lists 
itās not decentralized in the way they make it
it's closer to bittorrent
like yeah i could have 30 remotes for each contributors machine but thatās insane
soo.. there might be a moment where some project is unavailable?
because nobody around holding it
I don't remember the specifics enough to answer that 
sounds weird
most definitely
Git definitely is
but not commonly used that way, turns out that centralization is very convenient 
I think it's nice that radicle brings a lot of the convenience from GitHub and other forges except p2p
tbf this should be a non-issue in practice assuming you seed your own project at the very least
yeah
ye, if you care about the availability of your projects then you're gonna have to host them yourself
if you care about the availability of some other project then you can host that one too 
itās exactly like bittorrent 


they have some stuff for integrating issues and pull requests into the git repo itself too which iirc was kinda neat
though i don't remember how that works again
now I want to try Radicle again 
mabe tomorrow
there's a nixos module for it 
iirc there's one in KDE if you're using that

wait hyprland uses hyprlock doesn't it
i wouldnt know
maybe
worst case just create a oneshot service
I think it's 8

and the original was low frame rate, it's copying that

wait, I might of miscalculated... could be 4
a historical moment 
Redoing my documentation again... Personally system alone is 12 pages and that's mostly general architecture
No doxygen?
KDE 
no but you could probably write one that does whatever the bios setting code does to send the state back to the keyboard when you boot up.
it'll be somewhere in the USB HID standard somewhere.
order computer
can't pay with card bank transfer it in advance here's the pre-invoice
instant transfer but takes a business day to notice anyway
don't notice it for a day (probably will today, then send with delivery company
it's the worst snowy ice armageddon weather in a decade out there

bets are on: will i be able to look at the subathon from new hardware?
This is more the general evolving architecture not what every script does
This vid is such a funny contrast to gamers nexus absolutely tearing into the Intel keynote
Thanks to MSI for sponsoring this video! You can learn more about their new laptops at the links below:
Prestige 14 Flip AI+:
Learn More: https://msi.gm/SA913D2D
Pre-Order: https://msi.gm/S812A432
Prestige 16 Flip AI+:
Learn More: https://msi.gm/SE4D23B9
Pre-Order: https://msi.gm/S4CC76FB
Stealth 16 AI+:
Learn More: https://msi.gm/SF12D491
In...
They can both be true
"What does fork mean?" Oh no š
ćč¦č“ććććØććććć¾ćć
å®®åēä»å°åøć§2003幓ćććć¶ć¼ć©ć³ćććÆć¼ćććć¼ć©ć³ćććććććć¬ććć¹ć®ē¹ę®ććć¦ććę±ååć®ććŖć¼ćć¼ē“å¶ćććå°éåŗććććęćć§ćć
ļ¼åŗčćććć”ć®ćäøå±±åęåŗć®ććććę å ±ć親ćććć®ę„åøø...

how much is 3090 still worth for
the guy is a dev, right?
dude i didn't even use git and i know most of the terms especially the basic ones like fork
No no. dw the user doesn't know any coding and I actually explain that I need to manually editing the data on the code itself than just submitting the scores because it wasn't automated yet etc. I just assumed that user knows coding.
computers are hard. otherwise everyone would want one... oh.
i like stainless steel computers the most
the ringing in my ears after i headbutt them is enlightening

is part of cf dying for anyone else?
- gets verilog as a subject
- finally, i can learn it without procrastinating since it was on my backlog anyways of stuff to learn
- teacher is plain and utter dogshit, cannot teach for hell
why, must it always, be this way?
absolutely made the experience goatfuck
sometimes in life you just have to accept that you're going to have to learn it in another way
literally i have to yet again learn everything on my own this semester
didnt go so well last one (i failed mechanics)
and barely passed calculus and real mechanics 
I had a requirement to take the same genomics course from the life science department and from comp sci
I ended up doing the life sci one first and it was awful, it was almost entirely "do things this way, no shut up stop asking why don't think about it"

and I hated it. took the same course from the comp sci department and it was actually pretty fun
it was like, "use k-means clustering for this kind of gene lookup" "what is k-means clustering?" "don't learn what it means that's not important, also I have no idea"
what i learned about uni teachers at uni is that 75% of them are the autistic kid who really enjoyed math class and really hated reality, people, or the world, so they just want to be researchers and keep doing it but they -must- teach to hold the researcher position
it's not just teaching; a lot of researchers do progressively less and less research the more senior they get
and it becomes more about writing proposals for funding, not even papers

zoot zoot
nice GPU
damn, my mistyped word made me really troubleshoot this testing tutorial lmao, finally its working
3090, wowie
Wait if I'm remembering correctly you got something other than a cpu?
i supposedly should hunt 3 of those
but so far i can only find one
that one was $400, i ended up only paying $150 technically
after cashbacks and coupons and stuff from my card

cant believe it that it worked
im pretty stubborn with my standards
i only buy if i can get for way cheaper than its current market price
wrr
than certain someone who paid way more
for a 3050 level "workstation" card
wait no
even weaker than 3050
wtf?

help me set up cursor and claude caude (code) and git
no like
deadass
it was half the performance of a 3050
to verify if... web apps and mobile apps are working
dentge
like
what the hell
dentge
its bad that "trainable" made me think of dogs

wrrr
someone asking to do their homework is one of the most common things you see on freelance boards
if only they paid more than $18
i'd do it for them
truely a beautiful sight in the world of AI slop
cheating the old fashioned way
true
here they pay like $1-5 
lucky them
cheap service
i should look for apprenticeships to get some more free education 
hairdressing apprenticeship, rail engineering apprenticeship, coastguard apprenticeship, train driver apprenticeship
i'll be barbie 
reminds me of a person who would charge $300 to set up text gen webui
Hem? That thing's a plug and play type app though
is dropshipping software the new hustle...
cant be more profitable than making an AI b2b SAAS chatgpt wrapper startup using UGC marketing
nvm, it was $500
i mean no way he got the money... right....
if he somehow convinced rich as fuck oilers that his service is "top technical" work he might š
Where o:
š¤Vedal's sock puppet of himself is offering an apprenticeship?
Sigh neuro play despacito
try working for lawyers in patent cases, you can make thousands for doing what amounts to "look for a print statement in a codebase"
you'll just need to sacrifice some of your sanity for the "why" of what you're doing
wrr
people pay way, way more for way less, so I wouldn't be surprised
like, consider that people pay hundreds of dollars for geek squad to come and plug in a SATA SSD
$100 to transfer data to the new SSD, $100 to install windows, $40 to physically plug in the SSD
I'll explain to someone that it's really not much trouble and it'd be pretty simple to plug it in, but they'll say something like "no, I'm too scared to open my computer, I'll break it, I'd rather just pay"
people have the same sentiments about opening terminal/command prompt
their money 
not too far off 
tbf, people will make the same decision for electrical work or plumbing....although I'd rather flood my house before paying a plumber so
i just do dog
š
it show up
wow, apparently replacing a thermostat costs $100 to $500, not including the thermostat
along with 200 other dog emoji
for a job that really isn't too far removed from plugging something in
it is a gif
isnt that the same thing
idk
less theory
i think uni is a bit higher level ye
do u also have stupid topics like calculus and stuff
in the US, college refers to university, community college, etc.
in Canada, college is very distinct from university
i've been trying to figure out calculus for the last 3 hours but the concept of the antiderivative method makes no sense to me
so you know someone probably didn't go to school in the US when they say "uni"
i dont have calc but i have "math and physics" cuz gamedev needs some basic understanding of how physaics works and such
matrices also important
vectors and matrices probably the most important part
don't skimp on the linear algebra
it's basically the entirety of.....anything in 2D and 3D space
and ML and etc.
i was looking at the material from the school that i dropped out

there's sadly not really that much to do other than lots and lots of practicing to get the hang of it
gonna be a lot of staring at homogenous transforms, euler angles, and/or quaternions
i dont even understand linear equations i might actually be cooked
like y = mx + b?
what the hell is an antiderivcative?
f'(x) -> f(x)
integral
oh
aka integration yea
they taught you calc in middle school?
ye
its just derivative in reverse 
why tf do yall have that in middle school
basic calc but still calc
is this the thing where
middle school means something very different in some european countries
like gymnasium
ye
we didnt have it in school, only derivatives
middle school refers to secondary school which goes from ages 12 to 18 here
middle school in the US means ages 11-14 roughly
so there is no high school?
or there is school beyond 18 
bruh middle school here is 12-15

i always wonder if having worked for google was a mistake and i should have stayed in education instead
wtf 
this is our education system
yeah, so the european school weirdness
to me you're the fucking weirdos
swiss bruh
what in the fuck
we had integrals in 11th 
to put it more simply, we had integrals at age 17-18
ye same
ontario canada
in england, you have:
-> nursery (up to age 4-5)
-> primary school (up to age 10-11) (ks1, ks2)
-> secondary school (up to age 16-17) (ks3, ks4) <- derivatives and integrals here if you took further maths
-> college / sixth-form (up to age 18-19) <- derivatives and integrals here if you took maths
-> higher education at uni or an apprenticeship (with various degree levels)
In Canada, we don't have middle school. Elementary school -> 8 year. Then 4 year high school -> uni/collage
maybe i forgot its been so long ago 

god i wish i was in england

unc 
i literally know more english than german i can NOT write an essay in german y'all just need to learn 1 language
afaik for integration you just gotta get the area of the shit underneath the graph.
you do need to learn a language in secondary school, usually spanish, french, or german
this is similar system to my area in the US, but some states have middle school from grades 6-8, which is so weird to me
its up to you which one though
and, some schools dont do it, i believe
oh bruh i didnt know that
though you'd have to take an extra course to make up for it
my middle school was grades 7-8
is it also cosmetic there or you actually learn something
we started with learning 2nd language at age 5
the entire school was just two grades
the first 
school is a bad place to learn a language
like I said, shits weird with middle schools 
i also need to learn stupid french
cosmetic class
you learn cosmetics
But we don't have SAT or ACT or CRE
i still suck at french and ive been learning it for 13 years
why is there a hype train
where in canada? ontario? I went to grade 7-8 middle school there
type h rain
evil wants a record 
evil just stopped playing minecraft that's all
evil winning anything
all the standardized testing we have in the states sucks ngl, it's more learning how to take the test then understanding and applying the material
i dont get why people like to throw so much money at an aiš
we had standardized tests in canada but they were for school evaluation only
so like, what the issue?
you gotta do the riemann sum or something to get the area underneath the curve
hey, that's every student in canada who isn't in quebec

all I retained was je ne comprende pas francais
c'est la vie (it's used in english, it counts)
school is literally the ultimate slavery place where u have like 8h of school and then need to do homework for another 2hr
i didnt enjoy school but it wasnt too bad for me.
college is a lot nicer since you're not forced to do anything, but you gotta get that self-discipline
we had like 5-6h of school
and i didnt do hw so hw was 0h for me 
i prefer being an overworked googler trying to meet a deadline for a code project than school
well that's not a hot take
we have standardized testing too and itās exactly the same. learning how to take the exam itself
maybe standardized testing is the issue
if anyone currently in school is wondering whether the "I have an assignment due" nightmares ever stop, the answer is they do not
i only started doing my homework at home in late secondary school.
everything before that i jsut didnt do it or did it in class








cwt is cool









š“






