what to-do

lately, i've been feeling confused and anxious about what i want to focus on next. i feel like i can never escape the generalist schtick because there are too many areas i want to understand. at the same time, i want to nosedive into something like vllm and really get into the weeds—it’s been too long since i’ve contributed to an open project, and i want to do some proper coding. but it's a hefty time commitment before you can contribute something meaningful.

i've also been reaaallyy into reasoning models and want to run some rl experiments with grpo, specifically in combination with latent space thinking in an attempt reduce reasoning tokens. the large diffusion language models paper was also interesting—keen to play around with that. did they release the code for it? i can't remember.

to-do

• grpo experiments
• latent space reasoning
• open source contributions
  • vllm
  • unsloth
  • trl
• kernel programming; triton? cuda? or is this dead