asura
@asura.dev
they / any and will move your infra to GCP for food
Broke Blaise's (@wittywebhandle.bsky.social) brain after they claimed batteries weren't a form of vertical integration but openAI was. 🤪
Tech, pets, games, ... Linux, python, cloud
Be queer, do things
Broke Blaise's (@wittywebhandle.bsky.social) brain after they claimed batteries weren't a form of vertical integration but openAI was. 🤪
Tech, pets, games, ... Linux, python, cloud
Be queer, do things
You're framing everything in ChatGPT news bites and you shouldn't do that. That is not the definition of these technologies. Gemma3 and Olmo2 and even Qwen aren't going to tell me to sacrifice myself to space aliens.
Kagi I believe also uses transformers, too.
Kagi I believe also uses transformers, too.
November 26, 2025 at 6:36 AM
You're framing everything in ChatGPT news bites and you shouldn't do that. That is not the definition of these technologies. Gemma3 and Olmo2 and even Qwen aren't going to tell me to sacrifice myself to space aliens.
Kagi I believe also uses transformers, too.
Kagi I believe also uses transformers, too.
I'm not demanding you accept anything. I hate ChatGPT. You're not going to convince me to use it. You're also not going to convince me not to use it - since I have already made that decision.
You don't have to accept it. It simply will be. Another boring technology that goes right by.
You don't have to accept it. It simply will be. Another boring technology that goes right by.
November 26, 2025 at 6:34 AM
I'm not demanding you accept anything. I hate ChatGPT. You're not going to convince me to use it. You're also not going to convince me not to use it - since I have already made that decision.
You don't have to accept it. It simply will be. Another boring technology that goes right by.
You don't have to accept it. It simply will be. Another boring technology that goes right by.
Then why don't more people do it?
And you're specifically advertising for chatgpt right now.
Small purpose built LLMs don't generally do that.
And you're specifically advertising for chatgpt right now.
Small purpose built LLMs don't generally do that.
November 26, 2025 at 6:33 AM
Then why don't more people do it?
And you're specifically advertising for chatgpt right now.
Small purpose built LLMs don't generally do that.
And you're specifically advertising for chatgpt right now.
Small purpose built LLMs don't generally do that.
"Most" of them are offline.
People are only familiar with the largest frontier models.
If you really love chatgpt (you shouldn't) you can get offline chatgpt-oss.
Randomized probabilistic word association is what powers your Google search. They adopted Transformers over a decade ago or so.
People are only familiar with the largest frontier models.
If you really love chatgpt (you shouldn't) you can get offline chatgpt-oss.
Randomized probabilistic word association is what powers your Google search. They adopted Transformers over a decade ago or so.
November 26, 2025 at 6:33 AM
"Most" of them are offline.
People are only familiar with the largest frontier models.
If you really love chatgpt (you shouldn't) you can get offline chatgpt-oss.
Randomized probabilistic word association is what powers your Google search. They adopted Transformers over a decade ago or so.
People are only familiar with the largest frontier models.
If you really love chatgpt (you shouldn't) you can get offline chatgpt-oss.
Randomized probabilistic word association is what powers your Google search. They adopted Transformers over a decade ago or so.
Offline? And in your own language?
This is a very boring technology.
This is a very boring technology.
November 26, 2025 at 6:31 AM
Offline? And in your own language?
This is a very boring technology.
This is a very boring technology.
I'm going to mention again that I advocate for purpose trained small language models.
November 26, 2025 at 6:30 AM
I'm going to mention again that I advocate for purpose trained small language models.
None of these things require a PC or spoken language.
Telling me it doesn't require an LLM doesn't require bluesky.
yet here we are
Telling me it doesn't require an LLM doesn't require bluesky.
yet here we are
November 26, 2025 at 6:29 AM
None of these things require a PC or spoken language.
Telling me it doesn't require an LLM doesn't require bluesky.
yet here we are
Telling me it doesn't require an LLM doesn't require bluesky.
yet here we are
It's offline and can automatically translate into different languages, audio, or even braille. You get a bonus of accessibility. And it works offline. And it can be licensed from the very creators you just mentioned.
November 26, 2025 at 6:29 AM
It's offline and can automatically translate into different languages, audio, or even braille. You get a bonus of accessibility. And it works offline. And it can be licensed from the very creators you just mentioned.
You're right - it's not left.
The illustration there was that you can very loosely use words to get to the appropriate answer.
But you're right - add continental and it's California. Or add cnotninetal. Or connitental.
You should be able to follow the point in all of those mistakes.
The illustration there was that you can very loosely use words to get to the appropriate answer.
But you're right - add continental and it's California. Or add cnotninetal. Or connitental.
You should be able to follow the point in all of those mistakes.
November 26, 2025 at 6:28 AM
You're right - it's not left.
The illustration there was that you can very loosely use words to get to the appropriate answer.
But you're right - add continental and it's California. Or add cnotninetal. Or connitental.
You should be able to follow the point in all of those mistakes.
The illustration there was that you can very loosely use words to get to the appropriate answer.
But you're right - add continental and it's California. Or add cnotninetal. Or connitental.
You should be able to follow the point in all of those mistakes.
Offline? In braille? In 27 languages?
You all are going to be very surprised when you figure out the mistake that's being made here.
If I ask Gemini 3 which US state is in the mostest south left - it will say California. That is correct every time.
This is all that is needed to explain a stitch.
You all are going to be very surprised when you figure out the mistake that's being made here.
If I ask Gemini 3 which US state is in the mostest south left - it will say California. That is correct every time.
This is all that is needed to explain a stitch.
November 26, 2025 at 6:24 AM
Offline? In braille? In 27 languages?
You all are going to be very surprised when you figure out the mistake that's being made here.
If I ask Gemini 3 which US state is in the mostest south left - it will say California. That is correct every time.
This is all that is needed to explain a stitch.
You all are going to be very surprised when you figure out the mistake that's being made here.
If I ask Gemini 3 which US state is in the mostest south left - it will say California. That is correct every time.
This is all that is needed to explain a stitch.
I appreciate this answer :)
But books don't have ctrl+f or search boxes and about the time you've screwed up hours of work because you don't know what a certain stitch is... I mean at least vector search is useful. :)
But books don't have ctrl+f or search boxes and about the time you've screwed up hours of work because you don't know what a certain stitch is... I mean at least vector search is useful. :)
November 26, 2025 at 6:19 AM
I appreciate this answer :)
But books don't have ctrl+f or search boxes and about the time you've screwed up hours of work because you don't know what a certain stitch is... I mean at least vector search is useful. :)
But books don't have ctrl+f or search boxes and about the time you've screwed up hours of work because you don't know what a certain stitch is... I mean at least vector search is useful. :)
It doesn't need to.
If the training data is about crochet and you ask about a "double puff" it will provide some sort of answers about crochet - with the assumption that "double puff" has been used as a term it will produce the "correct" answer. With RAG updates it gets better without training.
If the training data is about crochet and you ask about a "double puff" it will provide some sort of answers about crochet - with the assumption that "double puff" has been used as a term it will produce the "correct" answer. With RAG updates it gets better without training.
November 26, 2025 at 6:15 AM
It doesn't need to.
If the training data is about crochet and you ask about a "double puff" it will provide some sort of answers about crochet - with the assumption that "double puff" has been used as a term it will produce the "correct" answer. With RAG updates it gets better without training.
If the training data is about crochet and you ask about a "double puff" it will provide some sort of answers about crochet - with the assumption that "double puff" has been used as a term it will produce the "correct" answer. With RAG updates it gets better without training.
Yes, you will.
The problem of hallucinations is inherent and unsolvable - and, in fact, a feature of LLMs.
But that does not mean you can't get deterministic answers. If you ask for a "double puff" a specifically trained crochet model is not going to go ChatGPT on you and talk about birds.
The problem of hallucinations is inherent and unsolvable - and, in fact, a feature of LLMs.
But that does not mean you can't get deterministic answers. If you ask for a "double puff" a specifically trained crochet model is not going to go ChatGPT on you and talk about birds.
November 26, 2025 at 6:12 AM
Yes, you will.
The problem of hallucinations is inherent and unsolvable - and, in fact, a feature of LLMs.
But that does not mean you can't get deterministic answers. If you ask for a "double puff" a specifically trained crochet model is not going to go ChatGPT on you and talk about birds.
The problem of hallucinations is inherent and unsolvable - and, in fact, a feature of LLMs.
But that does not mean you can't get deterministic answers. If you ask for a "double puff" a specifically trained crochet model is not going to go ChatGPT on you and talk about birds.
Is this an honest reply or kneejerk?
Are you saying you cannot make a small LLM that gives you correct answers? And you think they only output bullet points? I have news
A small model can be offline and edge - no need for 10K hours of video. You can get a step-by-step generated with licensed images
Are you saying you cannot make a small LLM that gives you correct answers? And you think they only output bullet points? I have news
A small model can be offline and edge - no need for 10K hours of video. You can get a step-by-step generated with licensed images
November 26, 2025 at 6:08 AM
Is this an honest reply or kneejerk?
Are you saying you cannot make a small LLM that gives you correct answers? And you think they only output bullet points? I have news
A small model can be offline and edge - no need for 10K hours of video. You can get a step-by-step generated with licensed images
Are you saying you cannot make a small LLM that gives you correct answers? And you think they only output bullet points? I have news
A small model can be offline and edge - no need for 10K hours of video. You can get a step-by-step generated with licensed images
Oh, on a side note: Your electric bill isn't *only* data centers.
Even though the US and EPA are destroying our solar plans as a country, battery and solar cost is through the floor.
As people implement this stuff for themselves your electric bill will go up in response to supply and demand. Sorry.
Even though the US and EPA are destroying our solar plans as a country, battery and solar cost is through the floor.
As people implement this stuff for themselves your electric bill will go up in response to supply and demand. Sorry.
November 26, 2025 at 5:49 AM
Oh, on a side note: Your electric bill isn't *only* data centers.
Even though the US and EPA are destroying our solar plans as a country, battery and solar cost is through the floor.
As people implement this stuff for themselves your electric bill will go up in response to supply and demand. Sorry.
Even though the US and EPA are destroying our solar plans as a country, battery and solar cost is through the floor.
As people implement this stuff for themselves your electric bill will go up in response to supply and demand. Sorry.
Given enough time we'll have open source and home made small language models that (yes, accurately, yes, without hallucinating) teach crochet and organic farming.
It's far too useful-yet-horribly-boring to ignore, just like blogs, patreon, wordpress, wikis, reddit, blah blah blah
It's far too useful-yet-horribly-boring to ignore, just like blogs, patreon, wordpress, wikis, reddit, blah blah blah
November 26, 2025 at 5:46 AM
Given enough time we'll have open source and home made small language models that (yes, accurately, yes, without hallucinating) teach crochet and organic farming.
It's far too useful-yet-horribly-boring to ignore, just like blogs, patreon, wordpress, wikis, reddit, blah blah blah
It's far too useful-yet-horribly-boring to ignore, just like blogs, patreon, wordpress, wikis, reddit, blah blah blah
Same as it ever was, tbh
Not long ago Wikipedia was the leftwing devil over Assault Rifle/Armalite Rifle brigades
If you're not behind solar and wind energy driven around by little AI/AI adjacent tech then you aren't actually left. Musk and Sama ruined stuff, but now they're gas/coal geezers
Not long ago Wikipedia was the leftwing devil over Assault Rifle/Armalite Rifle brigades
If you're not behind solar and wind energy driven around by little AI/AI adjacent tech then you aren't actually left. Musk and Sama ruined stuff, but now they're gas/coal geezers
November 26, 2025 at 5:46 AM
Same as it ever was, tbh
Not long ago Wikipedia was the leftwing devil over Assault Rifle/Armalite Rifle brigades
If you're not behind solar and wind energy driven around by little AI/AI adjacent tech then you aren't actually left. Musk and Sama ruined stuff, but now they're gas/coal geezers
Not long ago Wikipedia was the leftwing devil over Assault Rifle/Armalite Rifle brigades
If you're not behind solar and wind energy driven around by little AI/AI adjacent tech then you aren't actually left. Musk and Sama ruined stuff, but now they're gas/coal geezers
inb4 someone inevitably says something like "But eating beef is natural!" It's not. Beef marketing was a justification for refrigerated train cars to make the rich richer. xAI is the same thing.
People that think openai/xai are "the AI" still believe pork is the "other white meat"
People that think openai/xai are "the AI" still believe pork is the "other white meat"
November 26, 2025 at 2:15 AM
inb4 someone inevitably says something like "But eating beef is natural!" It's not. Beef marketing was a justification for refrigerated train cars to make the rich richer. xAI is the same thing.
People that think openai/xai are "the AI" still believe pork is the "other white meat"
People that think openai/xai are "the AI" still believe pork is the "other white meat"
Google sells TPUs and has for quite some time
They're only like 2-4 TOPS iirc but they're out there
gweb-coral-full.uc.r.appspot.com
They're only like 2-4 TOPS iirc but they're out there
gweb-coral-full.uc.r.appspot.com
Coral
Build intelligent ideas with our platform for local AI
gweb-coral-full.uc.r.appspot.com
November 26, 2025 at 1:38 AM
Google sells TPUs and has for quite some time
They're only like 2-4 TOPS iirc but they're out there
gweb-coral-full.uc.r.appspot.com
They're only like 2-4 TOPS iirc but they're out there
gweb-coral-full.uc.r.appspot.com
You don't have to stuff. Let other people that want to stuff do the stuff.
November 26, 2025 at 1:15 AM
You don't have to stuff. Let other people that want to stuff do the stuff.
Throughout history from Hellfire clubs to stag parties, the rich have been able to acquire everything from birth control, hormone replacement, abortion care. Bryan Johnson slurps up his son's plasma to make his peepee firmer.
In this thread: "We should be fighting a class war because I don't stuff"
In this thread: "We should be fighting a class war because I don't stuff"
November 26, 2025 at 1:14 AM
Throughout history from Hellfire clubs to stag parties, the rich have been able to acquire everything from birth control, hormone replacement, abortion care. Bryan Johnson slurps up his son's plasma to make his peepee firmer.
In this thread: "We should be fighting a class war because I don't stuff"
In this thread: "We should be fighting a class war because I don't stuff"