Meet

@protosphinx @DavidSHolz Nobel in medicine? Either this or that cancer drug

Updating my priors: If this is true, @DavidSHolz moves into the #1 spot among 21st century entrepreneurs. Above Elon and others.

@floinkus No ones going to scan themselves daily. Irrelevant cost comparison which doesn't make sense.

They say each scan is 806TB and they hope people are imaging daily. One scan/day × 806 TB × 365 days = 294,190 TB, or 294.19 PB. Using AWS S3 Glacier Deep Archive, which AWS lists at $0.00099 per GB-month / about $1 per TB-month, the storage-only cost is: 294,190 TB × $0.99/TB-month × 12 = ~$3.49 million per year, just to store your data

@etnshow Nightmare blunt rotation

BREAKING: Anthropic’s Dario Amodei, OpenAI’s Sam Altman, DeepMind’s Demis Hassabis and Mistral’s Arthur Mensch will meet for a 2-hour lunch today, per politico.

@KhaliqHussainnn I have a working product (VTON extension). Are you hiring? Do you have a company? or are you a recruiter?

We’re hiring 1-2 AI Engineers or Full Stack Developers Work - Building a website with an AI model that lets users upload a photo and instantly see how clothing brand products would look on them. Stack: Python, Computer Vision, GenAI, FastAPI, OpenAI, Cloud. DM resume + projects ( something similar to this ) #hiring #JobOpening #Ai #python

@narendramodi @EmmanuelMacron @G7 Kohli only drinks water from this region btw

Merci pour l’accueil à Évian, mon ami, Président Macron ! @EmmanuelMacron @G7

@antirez There is something called hard distillation. tough yes, impossible no.

Another important thing: Chinese models are not strong because they distill US models. Distillation of models via API is *impossible*. If somebody tells you the contrary, they don't understand machine learning:

@SarvamAI @hcltech @BessemerVP @khoslaventures @peakxvpartners Its time for SOTA 500B and beyond. godspeed🚀

We're thrilled to announce that we have raised $234M in the first close of our $300M Series B at a $1.5B valuation. @HCLTech and @BessemerVP have joined us in this round, alongside continued support from @khoslaventures and @peakxvpartners For countries and companies, sovereign control on the AI stack is no longer an optionality. Sarvam will be the partner of choice for this aspiration. The capital allows us to accelerate our momentum towards this full stack of models, compute, and deployments. A huge thank you to our customers, partners, investors, and the Sarvam team for your trust and belief in what we are building. We’re just getting started. Read more: sarvam.ai/announcing-ser…

@_Creation22 For 84L these questions are super easy. Any MLE can answer 80% of these. I was at least expecting multi head attention from scratch type of questions.

Interview experience at SARVAM , Bar is super high

@kmeanskaran Clean writeup🤝

@kunalvg No wonder.

It is disheartening to watch India’s IT veterans fumble AI this badly. They had the capital. They had the head start. They had every reason to see this coming. And even now, they are out here explaining why India should not build frontier models. Man, this just hurts.

My PR #37928 to JAX library got approved. - It fixes a sparse autodiff bug involving BCOO sparse arrays, vmap, and reverse-mode gradients. - In short: a batched sparse matvec gradient could fail because the cotangent shape didn’t match the original sparse data shape. - The fix uses JAX’s existing _unbroadcast helper to return the correct shape.

@googrish Multimodal fine tuning supported?

“don’t train your own model” is common ai advice. it's wrong. your token bill's the proof. today, we’re excited to launch castform into open preview. castform is the easiest way for you to train your own model, on your own data. open-weights models are performant and much cheaper. when trained on your task & proprietary data, they beat closed models. the thing standing between you and that was weeks of plumbing & years of ml expertise. with castform, model training is as simple as prompt engineering. @castformai bring your agent traces or raw corpora. castform turns it into training data, picks the right algorithmic recipes, manages gpus, and gives you an ide to watch and chat with your model as it learns. see what you can build with castform👇

@kingofknowwhere These all in 8 weeks of training? On top of their regular work? Too optimistic...

Just got off a call with someone from TCS. They want to train employees to build ai agents- 8 weeks of training. 160 hours in total. Their requirements are straight out of 1980. This is not going to end well.

@claudeai Enjoy this for a week before it's nerfed and becomes worse than the current one

Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use. Its capabilities exceed those of any model we’ve ever made generally available.

@kunchenguid Any metrics for spec driven development?

oh gosh, this is kind of a big deal DO NOT ASK YOUR AGENTS TO DO TDD! i now have empirical evidence that Test Driven Development is harmful for coding agents what other popular skills do you want me to debunk? details about the evaluation below 👇

@elonmusk @Coinvo @ShashiTharoor this seems right up your alley. Your erudite response is expected here.

It wasn’t. With rare exception, colonies were unprofitable, meaning more was spent building infrastructure like roads, railways, buildings, etc. than was exported. And look at places like Singapore and Hong Kong. Both were colonies for a long time and yet they are extremely prosperous.

ELON MUSK: "I don't believe Africa's poverty is caused by colonialism."

My PR #8220 was merged into @huggingface Datasets library. - It adds support for composed splits in streaming datasets - making split composition work more consistently between streaming and non-streaming dataset loading. A small but practical fix for ML data pipelines.

@CricCrazyJohns x.com/i/status/20633…

PICTURE OF THE DAY 🇮🇳 Rohit Sharma with Surya, Shreyas & MCA President Ajinkya Naik during T20 Mumbai league.

India's last 3 t20 captains: - Borivali (rohit) - Chembur (sky) - Chembur (shreyas)