analyticsaurabh's profile picture. http://tora.bohita.com
First Fashion Recommendation ML @ Rent The Runway 🦄, Founded ML at Barnes & Nobles. Past, @Virevol, Unilever, HP, ...

Saurabh Bhatnagar

@analyticsaurabh

http://tora.bohita.com First Fashion Recommendation ML @ Rent The Runway 🦄, Founded ML at Barnes & Nobles. Past, @Virevol, Unilever, HP, ...

And finally we can put that ridiculous notion to rest that workflows are agents

"... Loops, conditionals, and error handling can be done with familiar code patterns rather than chaining individual tool calls ..." This is why I was asking for types for MCP tool outputs; I lost count of how many times I was laughed in my face (and cannot even imagine how…



99% of companies investing in A.I tools have no idea what they are doing Most don’t even have a single person who has done ML

An M.I.T. study found that 95% of companies that had invested in A.I. tools were seeing zero return. It jibes with the emerging idea that generative A.I., “in its current incarnation, simply isn’t all it’s been cracked up to be,” @JohnCassidy writes. nyer.cm/FUZwzw8



word2vec remembers

It’s weird that before LLMs, it wasn’t even widely speculated that high-dimensional vectors encode Concepts, and the geometry of their relative positions encodes Meaning, in full generality. You’d think it’d have been a candidate hypothesis that was widely given 30% probability.



Noone complains about water usage

can someone explain why you’d ever put computers anywhere off planet? the only reason seems to me to be the far future one where we’re dumping too much waste heat into the earth system which is very far off



This is what NYC wants and I'm sure, the new mayor is smart enough to know which positions he will abandon

MASSIVE YIMBY VICTORY IN NYC All 3 NYC pro-housing measures pass by huge margins, clearing unnecessary red tape and making way for more housing!!

YIMBYLAND's tweet image. MASSIVE YIMBY VICTORY IN NYC

All 3 NYC pro-housing measures pass by huge margins, clearing unnecessary red tape and making way for more housing!!


It's funny when regressions happen in TORA but somehow, only I notice above is scripted, generated, edited by tora no post edits, no templates, no workflows...


Really cool Looks like a really well formatted Jupyter notebook :)

Introducing Codemaps in @windsurf! powered by SWE-1.5 and Sonnet 4.5 “Your code is your understanding of the problem you’re exploring. So it’s only when you have your code in your head that you really understand the problem.” — @paulg



Also ML

The only way to learn math is to do problems. All else is entertainment.



Anything from Alexia from GAN/GameModels/Tiny RNNs are required reading

New paper on how to fine-tune your Tiny Recursive Models!



The PhD vs not rift is still true today Yes there were exceptions. Mostly ossified now

You have to remember OpenAI in those days was lousy employer with not much pay nor any real career prospects for anyone. Most people thought they will shutdown soon after money would run out. People who joined were the ones who already were financially free, didn’t had to work…



Saurabh Bhatnagar reposteó

Recordings for the talk on "Tiny Recursive Models" are up. youtu.be/ETukUNsn_wQ

jm_alexia's tweet image. Recordings for the talk on "Tiny Recursive Models" are up.

youtu.be/ETukUNsn_wQ
jm_alexia's tweet image. Recordings for the talk on "Tiny Recursive Models" are up.

youtu.be/ETukUNsn_wQ

I will give a presentation on "Tiny Recursion Models" tomorrow at 1pm in the Mila Agora (6650 Saint-Urbain, Montreal). Its open to everyone, feel free to come by!



Saurabh Bhatnagar reposteó

So many teams in enterprise AI working on this and a single dude drops a paper just mogging them. Incredible.

A 7B model, tuned for forms and docs, beats giant models at pulling structured data. Beats GPT-4.1 on 1,000 extraction tasks, trained for $196. The team generated synthetic training data that preserves memory across chunks of a long file. That memory lets the model connect…

rohanpaul_ai's tweet image. A 7B model, tuned for forms and docs, beats giant models at pulling structured data.

Beats GPT-4.1 on 1,000 extraction tasks, trained for $196. 

The team generated synthetic training data that preserves memory across chunks of a long file.

That memory lets the model connect…


3 buildings in the coop here Absolutely mayhem

On the opposite side of the spectrum my kids got like 10 lbs of candy just going to different apartments at a 46 floor NYC building

bernhardsson's tweet image. On the opposite side of the spectrum my kids got like 10 lbs of candy just going to different apartments at a 46 floor NYC building


We dogfood TORA on bohita (the ai apparel portion) Hire your CMO, TORA to create and manage all ad campaigns for you. On brand, consistent characters, no post editing needed


Loading...

Something went wrong.


Something went wrong.