Balaji Varatharajan

3.3K posts

Balaji Varatharajan banner
Balaji Varatharajan

Balaji Varatharajan

@BalajiAI

ML nerd. Experimenting with life. Currently vibing at https://t.co/BqWEhjLINI

Present Katılım Ekim 2020
537 Takip Edilen1.8K Takipçiler
Sabitlenmiş Tweet
Balaji Varatharajan
Balaji Varatharajan@BalajiAI·
Today I’m excited to introduce VibeEdit - vibeedit.app It lets you edit any content - image, video, audio, just using text prompts.
English
3
2
7
1.3K
Nevasini
Nevasini@Nevasini1·
Been digging through STERA-10M all morning. some notes: 10M FPV samples for training world models, just released by @fpv_labs on Hugging Face. a few reasons I think this matters for embodied AI / spatial intelligence 🧵 huggingface.co/datasets/fpvla…
English
5
1
6
215
Balaji Varatharajan retweetledi
Sabba Keynejad
Sabba Keynejad@sab8a·
I have a confession to make... I accidentally built world's best subtitle API. Free @fal API credits for first 350 users who RT + comment "sub".
English
42
30
53
12.5K
Balaji Varatharajan retweetledi
JNS
JNS@_devJNS·
ZXX
213
1.3K
12.9K
543.1K
Balaji Varatharajan retweetledi
Deedy
Deedy@deedydas·
Billionaire Michael Milken joked “if a US company replaces the US-born CEO with a CEO born in India, I buy the stock” But he reveals he hasn’t backtested the idea. So we did. In the last 15yrs, that would’ve 50x’d your money: 7.5x more $$ and >2x IRR vs S&P500: 30% vs 14%!
English
562
1.7K
15.2K
3.6M
Balaji Varatharajan retweetledi
Jiayi Weng
Jiayi Weng@Trinkle23897·
Codex iterated a pure NumPy + cv2 closed-loop heuristic policy for VizDoom D3 Battle. No neural network training, no map, no object coordinates, no seed-specific routes. Just screen pixels plus public game variables, roughly the same signals a human player gets. It works surprisingly well. Notes and videos are now in the blog: #en-vizdoom" target="_blank" rel="nofollow noopener">trinkle23897.github.io/learning-beyon…
Jiayi Weng@Trinkle23897

Codex grew programmatic policies with no neural nets: max score on Breakout, and SOTA-level scores on MuJoCo. Maybe heuristics were not too weak. Maybe they were just too expensive to maintain. Maybe it's the next paradigm. trinkle23897.github.io/learning-beyon…

English
58
129
1.1K
2.8M
Balaji Varatharajan retweetledi
Indra
Indra@IndraVahan·
i turned 28 a few days back and one thing i realized is that none of the mutualfund-sip-savingsmaxxing discourse really mattered as much as i thought it did i did not have a linear 20s. apart from salary i made way more money through side gigs and weird internet opportunities. i can confidently say i made more in my 20s than 95% of people around me will in theirs. yet when i look back none of the money i saved feels nearly as meaningful as the things i talked myself out of buying or experiencing i could’ve bought that bmw g310rr at 22. i could’ve gone on that japan trip. instead i saved aggressively for a future where the same amount today barely covers 15 days of expenses i’m not saying saving or investing is useless. i’m saying most people vastly overestimate how much constant saving changes your life. your life usually changes because of one decision that exponentially compounds everything afterwards especially if you’re in your early-mid 20s on a non-linear path, don’t suffocate yourself obsessing over where every rupee goes
English
31
41
906
40.1K
Balaji Varatharajan retweetledi
Emma Steuer 🧚🤖
Emma Steuer 🧚🤖@emmysteuer·
You only live once, so make sure to spend as much time as possible on your computer. You won’t have access to it when you die
English
262
3.6K
30.5K
724.7K
Balaji Varatharajan retweetledi
Justus Mattern
Justus Mattern@MatternJustus·
People from top universities are great on average but nothing gets me more excited than talking to someone who went to a no-name uni (possibly in another country) and ended up at an org with a very high bar
English
25
77
2.4K
75.9K
Balaji Varatharajan retweetledi
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
1.5K
2.9K
23.1K
12.7M
Balaji Varatharajan retweetledi
Sann
Sann@san_x_m·
His name was Saurabh Kumar Ladda. He was 25 years old. Chemical Engineering from IIT Madras. MBA from IIM Calcutta. CFA Level 1 certified. He joined McKinsey and Company as a junior analyst in August 2022. On Friday, February 23, 2024, he returned home to his apartment in Wadala, Mumbai, after a work trip to Ahmedabad at 10:30 at night. At 11:15, he jumped from the ninth floor. In his last phone call to his girlfriend, he spoke about the pressure he was facing on his project. Police recorded statements from his flatmates, family, colleagues and seniors. The official police investigation concluded his death was due to pressure at his workplace. He was 25 years old. IIT. IIM. McKinsey. Every milestone India tells its young people to chase. Today is International Labour Day. The day the world stops to ask what we owe the people who work. India has no law that limits working hours for white collar professionals. The Factories Act caps factory workers at 48 hours a week. For salaried office employees, there is no legal ceiling. Saurabh Kumar Ladda is not the first. He will not be the last. Follow for real stories India never makes headlines about.
Sann tweet media
English
263
802
3.8K
624.9K
Balaji Varatharajan retweetledi
Charlie O'Neill
Charlie O'Neill@oneill_c·
I hate when I hear language model researchers complaining that they can't do any interesting research anymore because they don't have enough compute. Literally stfu. This attitude wrongly paints the world as infinitely less complex and fractal than it really is. Beautiful ideas do exist at scales <10B parameters. You can't pretrain the world's best model, who cares? Go and explore the unbounded space of ideas in the range of compute you have. The world is a microcosm and you can zoom in infinitely at any part of the scale to find interesting and useful things
English
21
16
283
18K
Balaji Varatharajan retweetledi
Henry Shi
Henry Shi@henrythe9ths·
Something strange is happening in tech. CTOs of billion dollar companies are quitting to take IC roles at Anthropic. Workday CTO -> MTS (Mar 2026) You[.]com CTO -> MTS (Mar 2026) Instagram CTO -> MTS (Jan 2026) Box CTO -> MTS (Dec 2025) Super[.]com CTO -> MTS (July 2025) Adept AI CTO -> MTS (Jan 2025) The mission is that real.
English
251
369
5.3K
2.6M
Balaji Varatharajan retweetledi
Divya Gandotra Tandon
Divya Gandotra Tandon@divya_gandotra·
For the first time I moved into a high-rise society, I was fascinated. Felt like this is what “upper middle class life” looks like. But after living here for 8 years, I’ve realised… it’s one of the biggest scams builders sold us. Tiny boxed flats stacked like cages, overpriced maintenance, artificial lifestyle… nothing but a pigeonhole dressed as luxury.
English
125
298
1.6K
264.1K
Balaji Varatharajan retweetledi
Polymarket
Polymarket@Polymarket·
BREAKING: AI-powered GLP-1 startup Medvi — built with $20,000 & two employees — is now on track for $1,800,000,000.00 in annual sales.
English
205
315
5.7K
2.5M
Balaji Varatharajan retweetledi
Michael Black
Michael Black@Michael_J_Black·
Tomorrow, I start a new full-time position as VP Digital Human Research at Epic Games. (And no, this isn’t an April Fools’ joke.) Today, I’m taking early retirement from Max Planck and will become an Emeritus Director. As an Emeritus Director, I will continue to supervise my remaining students, oversee ongoing projects, and wind down my department over the next couple of years, which is a normal process when a director retires. Being a Max Planck director is the best academic job in the world and it has been my great honor to co-found and help build the Max Planck Institute for Intelligent Systems. I love the institute deeply. But my statutory retirement age was looming and I’m not done yet. I’ve been working on capturing and modeling human movement for 30 years and it has gone from a fringe topic to something that works robustly and has immense potential. With AI today, scale matters, and achieving that scale increasingly requires industry. In particular, digital humans are moving from research prototypes to foundational technology across industries. At this point in my career, I want to get this technology into the hands of millions of users, while pushing the frontier of digital humans. I know that this is a challenging time for the games industry. Such times are precisely when people get creative, the industry is open to change, and real innovation can take root. Epic Games is the ideal place for this. The talent is deep, there is a compelling vision for how games will evolve, and the commitment is clear. I’m excited to join them and the rest of the Meshcapade team. There is no good way to express on social media the depth of my gratitude to the Max Planck Society, my co-directors, students, post docs, and staff. It has been an amazing 15-year journey because of you. Thank you.
English
55
44
1K
65.1K
Balaji Varatharajan retweetledi
bablu tiwari
bablu tiwari@realpurus·
This guy is my biggest inspiration cause how can you do martial arts, gym, chess, college, and work at one of the frontier AI labs in the world at the same time? Insane amounts of dedication and agency, nothing but respect to you brotha.
English
20
220
5.3K
280.6K
Balaji Varatharajan retweetledi
Tonino Catapano (tonnoz)
@ProductHunt just gave Product of the Day 🥇 to my silly vibecoded app I made for fun. Moreover, today marks exactly 10 days since SlapMac went live, which made me $20,700 in gross revenue. I'm still gasping at the screen 🫠 Many may think it's another absurd and lucky overnight success, but while it might seem that way, it wasn't. Let me explain. My story: Back in 2018 I was just a backend corp. guy who stumbled on @levelsio and his book MAKE, and something genuinely broke open in my brain. I wanted that life so badly that I went full obsessive: every @starter_story video, every @marclou marketing technique dissected and noted. Tweets from @tdinh_me were my inspo too. I started doing my small experiments (like @FpvBuddy): my playground to learn frontend, BaaS and the world of indie such as requesting payments, marketing & SEO. Then I started meeting people in the space: builders in Southeast Asia, then in Amsterdam. The kind of delusional, hungry optimists who bet everything on themselves. Those conversations changed how I think about making money online. Recently I even got my whole X profile roasted by @robj3d3, which was equal parts painful and clarifying haha, thanks dude. @transitive_bs also said something that stuck with me around that time: "what is blocking you from doing this full time. How can we make that happen?" I was about to give out with content making and tweeting when I met @did0f and his way of teaching tech through video format truly inspired me to keep going. And so, a few months later I made a video for fun, reviewing a repo that makes your laptop moan when you slap it, and things moved pretty damn quickly from there. 24 hours of hacking with the Hackadam crew later, there was a real app, written from scratch at the speed of light thanks to AI. Is SlapMac a business? No. But it's the kick that starts one. I'm done accumulating knowledge and ready to build for real this time. Overnight success is very often a lie. Mine took 6 years of learnings and a single slap 👋💨
Tonino Catapano (tonnoz) tweet media
English
149
48
1K
135.5K
Balaji Varatharajan retweetledi
clem 🤗
clem 🤗@ClementDelangue·
After @Pinterest @Airbnb @NotionHQ @cursor_ai, today it’s @eoghan @intercom publicly sharing that they’re finding it better, cheaper, faster to use and train open models themselves rather than use APIs for many tasks. And hundreds of other companies are doing the same without sharing. Ultimately, I believe the majority of AI workflows will be in-house based on open-source (vs API). It took much more time than we anticipated but it’s happening now!
clem 🤗 tweet media
English
81
185
1.7K
404.4K