Husain Zaidi's Bear Codex

I have been playing with my Jetson Orin Nano Super, trying to make it do things.

This is a polemic against an already disliked adjective.

Deepseek R1 has shown how Reinforcement Learning with simple rewards are enough achieve SOTA on benchmarks.

I have been consuming a series of really good sci-fi media since the start of 2024 which got me imagining and evaluating near-future scenarios.

LLMs like ChatGPT are built to generate words using their billions of weights trained over a significant portion of the crawled internet.

So much for my 2021 resolution.

My new year resolution was to make a project that achieves some sort of product-market-fit.