A Clojure port of XinJingHao’s PPO implementation using libpython-clj2, PyTorch, and Quil. PPO is a reinforcement learning method which has become popular because it addresses the problem of stability. The PPO implementation is tested using the inverted pendulum problem.

Bad code is usually more costly once you are beyond 100 lines of code.

My Windows gaming graphics performance dropped by 40% after some update. I already was using Linux for everything else except gaming for many years. So I tried out Steam on Linux and I was quite amazed how many games run on Linux via Proton. Just check out protondb.com for compatibility reports.

I did a bit of scripting trying to automate the TDD cycle given a task description. The problem is, that the LLM tends to jump forward and submit a full implementation instead of a minimal change. I guess the problem is, that LLMs are trained on complete solutions instead of minimal steps.

Yes, the main function is the worst part of the code. Whenever I discover a pattern, I refactor it and put it in a module.

Yes, I forgot. However Clojure multi methods are more generic in that the function computing the dispatch value can also be defined.

Good point. I definitely need to keep an eye on garbage collector pauses. I am using the new ZGC which supposedly has very short garbage pauses.

In the past I have done some rigid body physics in GNU Guile (see https://www.youtube.com/watch?v=zBq3kW2jVxs for example). Of course if you need to simulate many objects, you will hit performance problems sooner if you don't use C/C++/Rust. Also the developer of Jolt has solved quite difficult problems, so I was quite happy to use it instead of rolling my own.

I am aware of Neanderthal. I think it is more suitable for larger tensors and matrices. Will have to check out what ham-fisted is about.

I use a few macros for creating contexts (i.e. with-texture, with-stencil, with-scissors, with-tar). Also I have macros for rendering (onscreen-render, offscreen-render). However I try not to overuse macros.