Ojas Jain

@XGenWealth @BesuraTaansane @PhysicswallahAP *Bro after having 1 share

@BesuraTaansane @PhysicswallahAP @PhysicswallahAP Nice acting to gain support, I am investor in your company but looking at this video, I will exit my investment, You will regret for this video, for 10% rioters you are loosing 90% sincere students, You don’t look like teacher, Have some shame!

Alakh Pandey of @PhysicswallahAP scoring a spectacular self goal Dude is going to regret this big time - he forgot his company is listed

@harshitmul CFBR

@NightSkyToday Those are definitely the Simpsons throwing a wave in the second picture

JUST IN🚨: This is Most Detailed View of a Human Cell ever Captured!

Football won

@Stake Siuuuuuu

If any team scores in the second half, we’ll give away $10k to 20 random replies below 👇

Me and my after 3 bumps #worldcupmatch

My stan #WorldCupFinal

@mytechceoo More accurate title : “ LP catches VC Euromaxxing”

VC catches founder Euromaxxing

One of the most amazing things to happen this week from India !

@sama And that’s how the bubble burst kids

homeboy you're the one sellling public market investors on short-term space datacenters

No wonder sales people think Claude is shit for sales advice

@paulg Enterprise sales 1o1

Jessica decided 14 yo and I needed haircuts and made appointments for us. A traffic jam made us late, so only he actually got a haircut. When I got home I asked Jessica if she liked my hair, and she said yes, so everyone's happy.

I've been (re)watching CME295 from @Stanford since the past week, putting my notes on my website for reference. Link in the comments.

Since coding is super semantic & logical, perhaps another approach could also using different languages in the training paradigm ? Also, what coding languages would these be which are a focus for training ? Does this focus also change the model's performance on benchmarks meaningfully ?

glad sarvam is working on a coding model because you don't have a shot at a 'frontier model' if your training tokens are not >50% coding related and devops/swe style post training RL environments to add more agentic capabilities

@AnthropicAI @claudeai Please fix this bug

There’s no point in having sovereign AI that is not frontier

Here is the best path to building India's own AI lab : 1. Recruit HFT talent (both software, and some traders) from firms like TRC, JS & Optiver. CSE, Elec, Physics, Chip Designers, Mechanical Engineers for Thermal systems. The best of the best. 2. Match/surpass their in hand Quant salaries & surpass them with ESOPs. They might actually want to be a part of this due to more meaning and excitement around this field. 3. Get them together in Bangalore in a state of the art facility with all the creature comforts possible. 4. Give them unlimited resources & tell them material comforts will be assured, it's a national mission for them with full autonomy 5. Let them cook Not saying this is the only path out there, but it might get us there the fastest.

step 1) secure compute with public funding

@TVMohandasPai Inventing numbers, just like the ones you worship invent data out of thin air.

The most advanced AI models needs 50/60b $, huge investments in Hyper cloud, and many years! No country apart from US/ China can arbor do it like them. People who comment loosely need to understand this