I might not even properly start until next year, but I've been working out bits and pieces for a video rendering system. The concept is similar to manim, but with much lighter-weight dependencies, and oriented around general text and diagrams rather than offering a bunch of special-case support for mathematical figures. I also plan to have a sort of DSL that can be used to describe video content in a highly readable way (meant to be rigidly structured, but accessible by non-technical people and also a useful target for LLMs), itself using a new markup language of my own design.
Specifically about dependencies: one piece of the puzzle is a lightweight wrapper for buffers (such as the builtin `array.array`) to allow treating them like Numpy arrays (still with accelerated code for simple operations like elementwise addition). Video basically requires a 4d-array (length x width x colour x time). For drawing, rather than PIL I'm working with Cairo and Harfbuzz directly.
Building a spaced repetition / incremental reading / video / podcast / audiobook app. I wanted a better UX version of SuperMemo basically. Something I actually enjoy using myself to explore topics incrementally. It uses the latest FSRS algorithm (FSRS-7), and also supports SM-20, SM-18, SM-2.
I've started planning the update of user-facing help for my 23 year old magnum opus. Help probably hasn't been updated in 10 years (I prefer coding) and I'm rather dreading the tedium. No, I won't be using LLMs - not in my lifetime - it's hand crafted all the way.
starting of the year I really wanted to make any video game , and after a long time of struggling with javascript , I ditched it , learnt c++ and finally this month I made a ghost game inspired by flappy bird starting from scratch :)
I might not even properly start until next year, but I've been working out bits and pieces for a video rendering system. The concept is similar to manim, but with much lighter-weight dependencies, and oriented around general text and diagrams rather than offering a bunch of special-case support for mathematical figures. I also plan to have a sort of DSL that can be used to describe video content in a highly readable way (meant to be rigidly structured, but accessible by non-technical people and also a useful target for LLMs), itself using a new markup language of my own design.
Specifically about dependencies: one piece of the puzzle is a lightweight wrapper for buffers (such as the builtin `array.array`) to allow treating them like Numpy arrays (still with accelerated code for simple operations like elementwise addition). Video basically requires a 4d-array (length x width x colour x time). For drawing, rather than PIL I'm working with Cairo and Harfbuzz directly.
Building a spaced repetition / incremental reading / video / podcast / audiobook app. I wanted a better UX version of SuperMemo basically. Something I actually enjoy using myself to explore topics incrementally. It uses the latest FSRS algorithm (FSRS-7), and also supports SM-20, SM-18, SM-2.
https://en.wikipedia.org/wiki/SuperMemo https://github.com/melpomenex/plethora
I've started planning the update of user-facing help for my 23 year old magnum opus. Help probably hasn't been updated in 10 years (I prefer coding) and I'm rather dreading the tedium. No, I won't be using LLMs - not in my lifetime - it's hand crafted all the way.
starting of the year I really wanted to make any video game , and after a long time of struggling with javascript , I ditched it , learnt c++ and finally this month I made a ghost game inspired by flappy bird starting from scratch :)