rajan agarwal

@_rajanagarwal

making models really good at coding @openai

rajan.sh

Joined March 2018

Pinned
rajan agarwal
@_rajanagarwal
Nov 25, 2025
in order to have research agents that can run for days, we need context compaction i used RL to have LLMs naturally learn their own 10x compression! Qwen learned to pack more info per token (ie use Mandarin tokens, prune text) read the technical blog:
LLMs can invent their own compression - Rajan Agarwal
From rajan.sh
rajan agarwal
@_rajanagarwal
Aug 1
whether or not the models are “conscious” or not, or whether they can truly “reason” by definition, it’s becoming increasingly clear that pointing models & compute at interesting, hard problems can have incredible returns
Noam Brown
@polynoamial
Aug 1
An internal version of Astra, @OpenAI’s next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science. We believe it will be a major step for scientific reasoning. openai.com/index/ten-adva…
rajan agarwal
@_rajanagarwal
Jul 23
i remember negotiating with poke last april when someone told me to text this cool email bot as a power user it’s super cool to see how they’ve evolved & excited to see how they continue to grow!
Cognition
@cognition
Jul 23
Cognition is acquiring @interaction, the makers of Poke. If you've ever used Poke, you know why we're so excited.
00:00
rajan agarwal
@_rajanagarwal
Jul 23
last fall i interned at amazon agi working on rl behind Nova Act (CUA) genuinely some of the smartest and most supportive people i’ve worked with. learned a ton and super grateful that they took a chance on me any team would be lucky to have them
Polymarket
@Polymarket
Jul 22
JUST IN: Amazon cuts jobs from its artificial general intelligence team.
rajan agarwal
@_rajanagarwal
Jul 18
hillclimbing