Sandeep

fighting the demons in my head

[Sandeep @Sandeepvurity]

@theBuoyantMan Very difficult if you have to rent a decent flat.
Without that, yes damn easy.

[Sandeep @Sandeepvurity]

@SumitM_X There must be some value in them which others identified which we may not value much

[Sidu Ponnappa @ponnappa]

few seem to understand that AI slop is a problem caused by mediocre humans, not mediocre AI

[Sandeep @Sandeepvurity]

@FishAudio Fish

[Fish Audio @FishAudio]

Today we’ve raised $52M Seed and we are announcing the public launch of S2.1 Pro.

>It can clone a voice from 5 seconds of audio
>2x faster than Cartesia & 1/6th the cost of Eleven Labs
>most expressive model with word level control over emotion, intonation, pacing etc

We support frontier AI companies including HeyGen, LiveKit, Retell, Sanas, and OpenArt all run our model in production.

If you're a business and we can't cut your voice AI costs by 50%, we'll give you 1 year of Fish Audio for free.
Book a demo: https://t.co/vHkyZf9JoG

To celebrate our first birthday, we'll give you 1 month of S2.1 Pro for free. Like, retweet, and comment “Fish” to get it.

[Kimi.ai @Kimi_Moonshot]

Releasing the model weights and technical report of Kimi K3.

Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.

New model architecture: 2.5x the intelligence per unit of compute, not just more params.

Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.

Model weights: https://t.co/7m7eEg6Y0B
Tech report: https://t.co/yeu6cjpMCT
Tech blog: https://t.co/YTfiMSNM1f

[Sandeep @Sandeepvurity]

The next version models should be more cheaper than scoring more on swe bench scores

[Sandeep @Sandeepvurity]

20 years down the lane, the power of a country may be hugely dependent on its space tech.

[Sandeep @Sandeepvurity]

Recently moved to hyd, are there any tech groups in hyd?

[Sandeep @Sandeepvurity]

@jackedAJ @cursor_ai waiting for response of hyd’s event

[Sandeep @Sandeepvurity]

@jain_harshit People have been paying money for their substandard services which made it profitable.
You think Airtel spent 43,000 crore by making losses? Come on.

[Sandeep @Sandeepvurity]

Today realisation hit me that there is no problem which i couldn’t solve, somehow i get things done
Which made me think how much more ambitious i can be.

[Sandeep @Sandeepvurity]

Oh god, time to build something.
Thanks Team Codex 🙌🏻

[Elliot Arledge @elliotarledge]

If concepts didn't need to marinate in my brain for them to be fully understood, I would have learned everything that I know today, technically, about LLMs, deep learning, and CUDA programming in three months, give or take.

It's okay if you take a bit longer. Don't be to you hard on yourself.

[Sandeep @Sandeepvurity]

The next 8 days: nothing but the mathematics behind LLM architectures.

There’s a incomplete feeling when you know how to use something but can’t derive or explain why it works from first principles. That’s the gap I’m trying to close

[Sandeep @Sandeepvurity]

@alokbishoyi97 Dm

[Praveen Kumar Verma @Alacritic_Super]

Learn GPU Programming from these universities for free

Stanford CS149 – Parallel Computing
https://t.co/fCtyk7pYU7

Carnegie Mellon 15-418/15-618 – Parallel Computer Architecture & Programming
https://t.co/GaMlTFf7WM

California Institute of Technology CS179 – GPU Programming
https://t.co/lRmjaHv9Vp

University of Illinois ECE 408 – Applied Parallel Programming (CUDA)
https://t.co/PTxRujFwyC

University of Pennsylvania CIS 5650 – GPU Programming & Architecture
https://t.co/h4KRjUVeWU

NVIDIA CUDA Education & Training
https://t.co/mSTAHpYJFi.

[Turing Post @TheTuringPost]

A great source to understand or refresh Transformer architecture

It explains how transformers process text token by token, using self-attention to build contextual representations

Covers:

- Token embeddings and positional encodings
- The residual stream that carries information across layers
- Multi-head self-attention and long-range dependencies
- Feedforward networks, layer normalization, and residual connections
- Transformer blocks stacked into deep language models
- The language modeling head that predicts the next token

It also connects these concepts to GPT and BERT

[Sandeep @Sandeepvurity]

🛺 just hits different.

[Sandeep @Sandeepvurity]

Anyone who wants to get into AI.
Spend a month learning transformers paper. until you understand the why behind each and every word in it.