Long-form writing
Arguments and investigations that need more room.
I build programs around the true structure of a problem: exact enough for the task at hand, abstract enough that new capabilities become natural consequences of the system.
I’m developing these personal programs around gaps I see in research and processes I want to understand. Each is an ongoing experiment with a working result.
Personal Programs and experiments in development.
I’m developing two separate experiments that I plan to merge. One explores drawing directly inside Windows’ desktop compositor. The other is a live audio visualizer that maps stereo balance and loudness to luminous trails with lighting calculated from physical room and screen geometry.
The controller mixes a recurring baseline DPI with randomized sensitivity changes. This passive aiming experiment is designed to train hand–eye coordination through visual feedback across movement scales and strengthen later predictive aiming through learned visual priors.
This experimental Geometry Dash mod is being developed to learn decoration from copied levels through reinforcement learning. Its training pipeline pairs gameplay structure with decoration targets and conditions regional learning on music and nearby decoration.
This experiment works toward reinforcement learning on molecular structures. It asks whether the molecular geometries of dopamine and serotonin resemble the structure of musical frequencies more closely than they resemble that of controls derived from the same songs. Its comparison tools create those controls by shifting each musical frequency band in time. The final question: Do dopamine and serotonin show a larger similarity advantage for original songs over time-shifted controls than other molecules do when identical processing and scoring are used for every molecule?
This OSC control pipeline connects external signals to VRChat avatar behavior. The audio implementation uses computer playback as its input while the avatar defines how that signal drives shader properties and object behavior.
I build personal programs around the distinctions that change their behavior. The model then handles new cases rather than adding them as exceptions.
I test a premise through the consequences it produces. Their direction helps me decide whether that premise should continue shaping the work.
I fit each program to the work it must do. Every additional layer is justified by a demand that work actually creates.
Essays and project notes alongside shorter observations.
Arguments and investigations that need more room.
Project notes cover the methods and decisions behind ongoing work as well as its results and mistakes.
A short record of what I am working on and considering now.
A record of awards and releases alongside results from my work as well as talks and contributions.
Selected awards and formal recognition.
Launches and published work with measurable outcomes.
Speaking and teaching alongside mentoring and open-source work.
Email and profiles elsewhere.