A deep RL workshop weekend, school exploration for Ibhan, and a February wrap-up with strong step count, heavy sessions, and fresh PRs.
Explorations
I’ve generally been curious about Reinforcement Learning (RL) environments, and over the weekend I attended a workshop on Building Reinforcement Learning (RL) for LLMs from scratch.
The workshop was led by Sid and felt like a close gathering with free-flowing discussions. Apart from an overview of agent harnesses, RL environments, PPO, and GRPO, we also walked through Unsloth’s RL 2048 Game notebook.
I was already beaming with ideas around reliably scaling multi-agent and long-horizon systems.
Family & Friends
At home, we’ve been contemplating a school switch for Ibhan. We were looking for schooling that is more curiosity- and creativity-driven, while giving equal weight to co-curricular activities.
We visited VidyaShilp on Saturday and were happy to learn about their pedagogy and methods, which closely align with ours.
Workout
Got back to ~10k steps/day for the week and completed 4 heavy resistance training sessions.
It was also a wrap-up week for February, and I could see that I hit 48 new PRs (personal records) across most core exercises on Hevy, while averaging ~8k kg of training volume per session.