Devansh Tripathi

826 posts

Devansh Tripathi

Devansh Tripathi

@NotTheCh05en1

Indifferent to cynicism.

Bangalore, India Katılım Mart 2024
128 Takip Edilen78 Takipçiler
Devansh Tripathi
Devansh Tripathi@NotTheCh05en1·
@alex_whedon congratulations it'll be interesting to see how it manages to avoid the quality cliff on a long context window
English
1
0
9
13.6K
Alexander Whedon
Alexander Whedon@alex_whedon·
Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.
English
643
1.2K
8.9K
5M
Sam Altman
Sam Altman@sama·
pretty excited for voice models to get great its interesting to watch how people are already starting to change the way they interface with AI
English
871
227
5.9K
521.6K
Devansh Tripathi
Devansh Tripathi@NotTheCh05en1·
@himanshustwts cascading architectures have already become reliable(fast) enough to be deployed for enterprise use cases S2S models are still not auditable on a large scale for adoption to kick in, but I reckon they are have lot more room for "taste" than a cascading one
English
0
0
0
97
himanshu
himanshu@himanshustwts·
based. i think there are two very interesting (and hard) problems the frontier is yet to shape on > scaling efficient multimodal + multi agent orchestration > scaling taste (on specific domains). as a vague example, creative writing which is the very first principles of text based language modelling has a lot of scope to improve. again its subjective as it boils down to the nuance of taste, yet to be improved from slop.
Sam Altman@sama

pretty excited for voice models to get great its interesting to watch how people are already starting to change the way they interface with AI

English
5
2
46
3.3K
Humi
Humi@byteHumi·
never in my life I have used such a horrible platform like LinkedIn I wanted to buy premium on this platform (for some reasons) took the highest premium plan and instantly after that they banned my account saying "I am impersonating someone" half of the time this platform doesn't work ..same glitches all the time.. you can literally check my name on the card I am paying with and match it with the profile but no let's took the money and ban him instead 85% of your platform is ai bots posting ai slop all the time If not for the good people out there..there's no point of using this crap.. urging people to be on X anyways they finally after 7 days realized I am a real human being
Humi tweet media
English
15
1
52
1.5K
Pakistan Railways
Pakistan Railways@PakrailPK·
پاکستان ریلویز سے سفر کیجیے، یہ محفوظ ہے
Pakistan Railways tweet media
اردو
571
1.2K
7.9K
306.6K
Devansh Tripathi
Devansh Tripathi@NotTheCh05en1·
Me : Yeah, CAC is proving to be a headache My gf who literally designs missiles :
GIF
English
0
0
2
127
Devansh Tripathi
Devansh Tripathi@NotTheCh05en1·
I went to therapy because I thought it would help me reconcile with my fucked up childhood. The marginal utility decreases significantly after a couple of sessions, and in the long run, it dulls your ambition in the garb of self awareness. Best therapy is becoming capable (for a man)
English
0
1
2
155
m
m@yowot_·
the love of ur life is a guy on twitter with 10 followers and weirdly mysterious profile picture
English
654
565
6.3K
245.2K
ib
ib@Indian_Bronson·
I thought everyone was joking about going to therapy but apparently it’s a multi-billion dollar industry and loads of people are just raking it in listening to idiots vent about stuff
English
6
6
209
6.4K
World of Statistics
World of Statistics@stats_feed·
🇮🇳 Births per woman in India. 1960: 5.92 1980: 4.78 2000: 3.35 2020: 2.05 2025: 1.94
English
204
204
4.7K
225.9K
Lex Blazer 🇨🇦
Lex Blazer 🇨🇦@LexsWorld·
@stats_feed The one country where we really gotta get that number down to <1.00 or they'll soon outnumber us all.
English
10
1
655
9.2K
Omar Abdullah
Omar Abdullah@OmarAbdullah·
Bloody hell 😱😱😱!!
English
1.8K
1.6K
22.9K
2M
kkloouttgg
kkloouttgg@ThcdThcd113778·
@devahaz @Molson_Hart Shut the fuck up pajeet . We are taking high paying jobs here your kind get using fake certificates and nepotism
English
2
0
14
346
molson 🧠⚙️
molson 🧠⚙️@Molson_Hart·
I will explain how American citizens can beat an H1B. First, there are 3 obvious ways, then I will teach you how to really beat them. 1. Work for a company that doesn't hire H1Bs, from startups to family owned companies 2. Start your own business 3. Leave the country for somewhere cheaper and work US jobs remotely Okay, here's how to really do it. Find major cash-rich corporations that hire H1Bs. Apply for their jobs. Get rejected. Find the allegedly secret job postings they do. Apply for those. Get rejected. Offer to work for the same salary as the H1B, note how many people they are hiring who are citizens vs. non-citizens and whether or not the companies have in-group preference (I am constantly told that Indians only hire Indians). Once you have stacked enough rejections (because people say it's so impossible to get hired because h1b's...), cycle through labor attorneys who will work on contingency and sue these corporations for discrimination against citizens for H1Bs. You don't need to take these to trial, just settle for cash over and over. That's how you beat H1Bs.
molson 🧠⚙️ tweet media
Hobbes the Cat@Bannedforself

@Molson_Hart LOL. Molson thinks he can "beat" a H1B. So, you do understand that the US company or university doesn't actually want a US citizen for the job. They put out newspaper ads with fake emails. The H1B ad is real (sorry no green card or citizen) Explain how you will beat this.

English
82
278
1.8K
124.9K
Retep
Retep@C137ick·
@NotTheCh05en1 @moodslayer95 Arabs Iran Iraq turkey all mogg india In terms of wealth looks height education food PCI HDI female safety and education
English
2
0
0
180
doomslayer
doomslayer@moodslayer95·
I judge people by their enemies and it's striking that Indians don't have a single decent or successful enemy
English
12
18
194
9.4K
Retep
Retep@C137ick·
@moodslayer95 Whites, chinese, muslims aren't successful?
English
4
0
0
1.2K
Indian Tech & Infra
Indian Tech & Infra@IndianTechGuide·
🚨 India’s Top Workplaces for 2026. 1. Infosys 2. Accenture 3. Amazon 4. JPMorgan Chase 5. SAP 6. IBM 7. Fidelity Investments 8. Alphabet 9. EY 10. Wells Fargo (LinkedIn)
English
137
165
3.2K
142.5K
Devansh Tripathi
Devansh Tripathi@NotTheCh05en1·
I am one of those 150M who lives in those "glistening" enclaves. A big way, the upper class copes with being indifferent to the reality is by constantly underestimating/underplaying their own privilege. Similarly, a lot of my Norwegian friends have a habit of downplaying their own agency when it comes to enacting change. Almost as if the collective Norwegian consciousness somehow believe that they don't deserve their riches/QoL
Devansh Tripathi tweet media
English
0
0
6
742
ib
ib@Indian_Bronson·
Norwegians don’t desire their children being raped by migrants but they don’t object to it. To explain: In India, 10% of the country (~150M people) have 70% of the wealth, while the bottom 50% (~725M) hold 15%. So the upper class lives *extraordinarily* well, in glistening enclave high rises in a country with space program. The bottom 50% live in a huge, extended series of slums with starvation and open air defecation. The upper class in India, bigger than many countries’ entire populations, don’t have to deal with any of that and so across (many) religious & ethnic lines just don’t care. But what you can then say accurately say about Indian society is it just doesn’t care that much about the squalor of total poverty. It simply doesn’t mind poor people shitting in some street, because it’s not *theirs*. If they did, as a society, they’d stop that. Similarly, if it’s their own daughter being raped by a migrant, Norwegians won’t like that. But if it isn’t their very own, it’s just not a big deal to them. They don’t object.
Remix News & Views@RMXnews

🇳🇴 A Syrian migrant found guilty of raping a 13-year-old Norwegian girl in a bike shed in Tromsø will serve just six months in prison after the sentencing court ruled that the attacker had a "low IQ" and "a reduced understanding of reality." Abdelmonem Abdelrazak Al-Yousef was found to have raped the girl in September 2024, less than a year after arriving in Norway from Syria. He met her in Tromsø city center before assaulting her on a park bench and leading her to a secluded bike shed near the Harbour Terminal where he raped her. He initially denied meeting the girl, but after semen was found at the scene and traced to him, he admitted to meeting the victim but not engaging in sexual activity. Psychiatric assessments showed he had a very low IQ, and the court ruled that "the defendant is most likely no further along in development than the victim." A change to Norwegian law last year removed the minimum prison sentence of three years for rape, allowing the court to impose a partial suspended sentence conditional on probation, meaning he will only serve six months inside.

English
3
14
188
23.6K
EP
EP@eptwts·
@NotTheCh05en1 well, digging through a book for hours is just a low ROI task compared to actually working imo i'll take the tradeoff of using an LLM as a qualifier instead of wasting time that can acc get me an ROI
English
2
0
8
587
EP
EP@eptwts·
i think nowadays with access to LLM's & deep research agents, reading books is inefficient when it comes to acquiring info... you can get an agent to go through 1000's of books & extract the most applicable info to your specific case without having to waste hours reading if you do it for leisure then that's fine, but reading as a way to "acquire knowledge" is one big cope pushed by the self-improvement cult
Nina@nina__369

@eptwts you read a lot of books?

English
90
13
395
85.7K
Devansh Tripathi
Devansh Tripathi@NotTheCh05en1·
because it doesn't include the loop of discovering something and rejecting it, which ultimately builds taste. you are basically outsourcing the judgment of what's required in your life to an LLM. If you have other avenues to exercise it from time to time, that could be a way to go, but I don't think that's a healthy long-term practice that one should partake in.
English
2
0
16
593
EP
EP@eptwts·
@NotTheCh05en1 and why can't this be done with an extracted list of concepts that the LLM deems best applicable to my life? it is just reading a curated list after all
English
3
0
9
1.7K