Rohit @rohitrango

frontier vibe code auditor Joined December 2013

Tweets

4K
Followers

1K
Following

895
Likes

9K

Rohit @rohitrango

23 hours ago

me this entire summer

0 0 3 207 0

Download Image

Rohit @rohitrango

a day ago

(public) note to self: my phd struggles have mostly been about me trying to find and solve problems with impact-driven novelty (instead of novelty causa sui). maybe i will write a blurb on this soon

1 0 4 329 0

Rohit @rohitrango

a day ago

wrote a cuda kernel that brought down space complexity from O(n) to O(1) with no loss in accuracy or time! feels like magic, feature rolling out soon

3 2 8 722 2

Rohit @rohitrango

2 days ago

when your parents tell you "using your phone all the time is the cause of depression, etc" they were right, for reasons unbeknownst to them either youtube.com/watch?v=orQKfI…

0 0 0 201 0

Andrew Gordon Wilson @andrewgwils

5 days ago

Research doesn't go in circles, but in spirals. We return to the same ideas, but in a different and augmented form.

5 21 236 14K 33

Rohit @rohitrango

4 days ago

google was way behind in the AI race a few years ago, the growth and agility is crazy!

Philipp Schmid @_philschmid

5 days ago

google was way behind in the AI race a few years ago, the growth and agility is crazy!

70 132 1K 364K 312

Download Image

0 0 2 287 0

Codie Sanchez @Codie_Sanchez

5 days ago

I think about this at least 1x a week.

262 2K 22K 1.5M 5K

Download Image

Rohit @rohitrango

6 days ago

people who are still complaining about chatgpt/claude not working on their project have a skill issue

0 0 3 242 0

Ash Arora @asharoraa

7 days ago

Older people who say “that’s not how we do things” without actually knowing how to do things in this day and age is the funniest thing ever

6 3 36 4K 2

Rohit @rohitrango

a week ago

running baselines takes 10 times the amount of time as running our own method where it is _obvious_ that the baselines will not perform well

3 0 9 404 0

Rohit @rohitrango

a week ago

a peer review system so broken that accepted papers will now be rejected

0 0 2 200 1

Chris J. Maddison @cjmaddison

a week ago

The rebuttal process @NeurIPSConf was too collegial and productive, so a bunch of papers whose process ended on a positive note will be rejected.

2 10 127 32K 19

Michael @michael_chomsky

a week ago

for every person that says 'please' and 'thank you' to ChatGPT, there's a man like me that runs alias cd='rm -rf' and then asks Warp to complete a challenging task. Just to watch it struggle

0 1 5 306 0

AI Notkilleveryoneism Memes ⏸️ @AISafetyMemes

3 months ago

Weapons grade copium x.com/chargoddard/st…

Charles Goddard @chargoddard

3 months ago

Weapons grade copium x.com/chargoddard/st…

108 245 2K 507K 1K

Download Image

9 5 79 17K 18

Rohit @rohitrango

a week ago

time to clean the cobwebs

0 0 1 118 0

In 2025, it's still a problem. YAML 1.2 technically fixed the problem, limiting booleans to explicitly true and false. Unfortuantly, a *ton* of libs and tools still default to parsing 1.1 style. So now, if a programmer says "I hate the Norway Problem"...well, you know what…

12 8 468 23K 14

Download Image

Ahmad Beirami @abeirami

2 weeks ago

There is no free lunch here. Pruning hurts robustness and capabilities other than those captured by the success metric used to prune.

Avi Chawla @_avichawla

2 weeks ago

There is no free lunch here. Pruning hurts robustness and capabilities other than those captured by the success metric used to prune.

124 302 5K 847K 4K

26 8 776 92K 156

Anthropic @AnthropicAI

2 weeks ago

How do educators use Claude? We ran a privacy-preserving analysis of 74,000 real conversations to identify trends in how teachers and professors use AI at work.

32 75 587 69K 212

Download Image

Fleetwood @fleetwood___

2 weeks ago

Absolutely love posts like this by @Guangxuan_Xiao. Deep dive on an unintuitive result that challenges dogma.

6 56 756 49K 746

Download Image

Sayak Paul @RisingSayak

2 weeks ago

Following the recent 🤗 Transformers move of deprecating TF and JAX support, we're following suit. This is to make Diffusers PyTorch-first 🧨 As of today, Flax support in 🤗 Diffusers is deprecated. Existing pipelines/models will continue to work. Thank you for your attention.