Prityush bansal

793 posts

Prityush bansal

Prityush bansal

@Prityush

Just another AI/ML Enthusiast

Gurgaon Katılım Haziran 2024
595 Takip Edilen168 Takipçiler
Prityush bansal retweetledi
Gareeb Scientist
Gareeb Scientist@gareebscientist·
I visited OLA's Cell factory, saw how the cell is made and interview/converse their Factory manufacturing lead, cell R&D lead and COO with tech questions .... the conversation was pretty interesting to say the least. You can make your own opinion. I was impressed. Link below.
English
74
261
1.8K
70.6K
Prityush bansal
Prityush bansal@Prityush·
Something interesting I observed, gpt 5.5 and claude opus 4.7 seem to be inverse of their previous counterparts What do I mean by that? The series before claude opus 4.7 (opus 4.6, 4.5) were good for vibe moonshots, describe an idea and it will let you build a lot of things, will it work, probably not, it tried to combine multiple things and be optimistic towards its goal/prompt. But with 4.7 its much more realistic, it pushes back well, removes the unknown parts, and tries to ground itself in how the thing will work. If any plan given by opus 4.6 had a 30% chance of working, with opus 4.7 its more like 70-80%, thats progress But gpt 5.5 has had the inverse effect. Earlier versions(5.2, 5.3, 5.4) were much more grounded and had 70-80% chances of working, gpt 5.5 is now doing vibe moonshots, it goes deeper and comes out with a plan that has a 30-40% chance of working This is the inverse effect For a long time, the alpha was claude working as a dev and gpt used to be critical of that work and delete the random stuff claude put in, and this combination worked well for any project you were doing. Now gpt working as a dev with claude being critical of the work works as an alpha. Same way of working, just the models have changed their roles. I am not particularly sure how this came into play, just sharing what the best ways of working with these models are We believe in moonshots, so shifted from the max plans of claude to gpt
Prityush bansal tweet media
English
0
0
1
65
Prityush bansal
Prityush bansal@Prityush·
The country that became the world’s back office now has to decide whether it becomes the world’s AI operations layer or just gets optimized out of the old model
English
0
0
0
24
Prityush bansal retweetledi
GalaxEye
GalaxEye@GalaxEye·
Separation Confirmed! The world's first OptoSAR Satellite is now in space. Made in India for the world. Go Drishti! Go @GalaxEye! Go India!
GalaxEye tweet media
English
204
1.2K
5K
197.3K
Prityush bansal
Prityush bansal@Prityush·
The new software hierarchy: shape rotators build systems wordcels name systems vibe coders break systems founders call it a platform
English
0
1
1
64
Prityush bansal
Prityush bansal@Prityush·
@DamiDina To be fair i asked chatgpt for ragebaits, seems like it didnt understand me
English
0
0
0
26
Prityush bansal
Prityush bansal@Prityush·
A lot of people do not want to build AI They want the social status of understanding it
English
4
1
10
228
Kaalye
Kaalye@bhairavum·
@Prityush It’s ok if ai understands us no?
English
1
0
2
29
Prityush bansal
Prityush bansal@Prityush·
The future of work is me explaining the same thing to ChatGPT, Claude, Cursor, and one intern. None of them understand. The intern is improving fastest
English
0
0
1
34
Prityush bansal
Prityush bansal@Prityush·
The main thing AI has automated is confidently missing the point
English
0
1
1
50
Prityush bansal
Prityush bansal@Prityush·
The average AI startup today is a pitch deck with latency
English
1
1
2
75
Prityush bansal
Prityush bansal@Prityush·
AI is the new MBA A socially acceptable way for people with no technical depth to sound inevitable
English
0
0
1
39
Y Combinator
Y Combinator@ycombinator·
This summer we're launching YC Paper Club to bring researchers & builders together. Every few weeks we'll host a small group dinner in Mountain View to share and discuss new research papers. Interested in both research & building the future? Sign up at: events.ycombinator.com/ycpaperclub
English
27
40
508
116.5K
Vaibhav Sisinty
Vaibhav Sisinty@VaibhavSisinty·
AI models just got their first neuroscientist. 🤯 His name is Silico. For decades, every AI model has been a black box. Even the engineers who built it could not explain why it worked. You ship the model. You hope it behaves. You can't open the skull. Goodfire just opened the skull. Silico decomposes a model's brain into interpretable features, diagnoses hidden failures before they ship, and lets engineers steer behavior like editing code. The same interpretability stack already discovered a new class of Alzheimer's biomarkers. The neuroscientist for AI just helped the neuroscientists for humans.
Goodfire@GoodfireAI

Introducing Silico: the platform for building AI models with the precision of written software. Silico lets researchers and engineers see inside their models, debug failures, and intentionally design them from the ground up. Early access is open now. 🧵(1/10)

English
4
7
55
8.3K
Prityush bansal retweetledi
M a k u b i
M a k u b i@krissmakubi·
@stripe @link Me in the morning asking agents what they just purchased.
English
0
1
48
2.8K
Suyash Singh
Suyash Singh@thesuyashsingh·
Countdown begins! Stay tuned for the Mission Drishti Launch. May 3rd, 2026 12:29 pm IST galaxeye.space
Suyash Singh tweet media
English
27
56
338
17.9K
Sagorika
Sagorika@sagorika_s·
I would be so embarrassed shilling Communism, man. This is just the absolute worst. The cognitive dissonance needed to follow and appreciate this should break brains, but sheep barely think.
Abhishek@AbhishBanerj

Dhruv Rathee showing his commitment to environment 1. First class on Emirates 2. First class suite on Singapore Airlines 3. Business class on Qatar Airways Please save trees in Nicobar Someone has to offset carbon footprint of Dhruv Rathee flying in first & business class!

English
6
24
179
3.3K
sankalp
sankalp@dejavucoder·
claude sonnet 4.8 wen to reverse the damage caused by opus 4.7 lol
English
7
0
68
3.9K