Neil Shah
@neilhshah15
Followers
666
Following
1K
Media
23
Statuses
843
CS + Math @Harvard. MARL @KempnerInst and CSPs @HarvardStats. Always @thecrimson.
Cambridge, MA
Joined January 2023
only found out GitHub is down because the copy code button wasn’t working on a file that I mistakenly deleted why is the copy code making a backend call lol
0
1
2
This is the biggest performance delta we’ve seen since launching Design Arena Gemini 3.0 Pro has taken #1 overall and #1 in 4 of our 5 code arenas - Website, Game Dev, 3D Design, and UI Components Well-earned congratulations to the @GoogleDeepMind team on a remarkable
56
139
2K
bullish
GPT-6 will be renamed GPT-6-7, you're welcome
1
0
2
villain origin story
0
0
3
We’re super excited to introduce DIRT: The Distributed Intelligent Replicator Toolkit: https://t.co/RidXwgGriR DIRT is a GPU-accelerated multi-agent simulation platform enabling artificial life research in dynamic, open-ended environments at unprecedented scales. 1/n
2
6
16
Our latest post explores on-policy distillation, a training approach that unites the error-correcting relevance of RL with the reward density of SFT. When training it for math reasoning and as an internal chat assistant, we find that on-policy distillation can outperform other
63
398
3K
i kinda like popup bagels — and am a bagel fanatic, as any of my friends could easily tell you — but that doesn't change the fact that this piece by HGL in this week's @crimsonFM made me giggle: https://t.co/ZZIMuNMm7f
thecrimson.com
This is not a review of PopUp Bagels. | Magazine | The Harvard Crimson
1
0
2
it took me 10 years of using LaTeX to realize you didn't need curly braces after \hat
1
0
2
most underrated feature of macOS 26 is that it no longer takes O(seconds) to insert an emoji
0
0
2
did you mean... π-thon?
Python 3.14 stable dropped today! Congratulations to everyone involved. You can install it now with `uv python upgrade 3.14`.
37
652
8K
what it’s like working for your college newspaper for four years
"Not all jobs are supposed to earn you a living wage" oh for sure, this is my recreational job.
2
13
142
the original VCBench
This paper claims LLMs are better at selecting successful founders than VCs "We introduce VCBench, the first benchmark for predicting founder success in venture capital (VC)" "most models surpass human benchmarks"
9
12
223
ICYMI: HAPPY 3,4,5 ALL-SQUARED DAY!! Today's date, 9/16/25, is a trio of squares (in a perfectly squared year, no less!) – and if that weren't exciting enough on its own, those squares correspond to the Pythagorean theorem applied to a 3,4,5 right triangle. Many of us saw the
8
8
60
Excited and humbled by this feature from the @KempnerInst @Harvard: Leveraging Modern AI to Catalyze a New Era in Naturalistic Computational Cognitive Science https://t.co/qHnGjG5b3P
kempnerinstitute.harvard.edu
To better understand the mysteries of human cognition, scientists at the Kempner Institute are turning to a powerful new ally: artificial intelligence. Kempner Research Fellow Wilka Carvalho and his...
0
5
33
super cool stuff from a super cool person!!!
Three weeks ago, we started building an AI game engine. But some models kept making things look... sloppy. So we turned finding the best one into a game. In three weeks, that game grew to 35K+ users across 135 countries. Introducing @designarena_ai, the fastest-growing
0
0
0
Excited to share a new project spanning cognitive science and AI where we develop a novel deep reinforcement learning model---Multitask Preplay---that explains how people generalize to new tasks that were previously accessible but unpursued.
2
8
38
77 years ago this week Claude Shannon ushered in the field of information theory with his paper "A Mathematical Theory of Communication," which has been cited over 100,000 times: https://t.co/jdJaBSIZ1Q
19
311
1K
preprint on research direction we're calling "Naturalistic Computational Cognitive Science" below is a summary of our argument. if you disagree with us, please let me know! feedback very welcome https://t.co/3KUeWvzzEk
5
10
41