Arissmiller.net / Field notes

notes_on_ai_usage.md
Preview

Notes on AI Usage

This site and these prototypes were created with ChatGPT codex. This was my first time experimenting with AI Coding tools and it was really exciting. After spending some time creating software with these tools, I've found they are very valuable for prototyping, but it is very easy to get over your head especially when designing complex systems that are outside of your knowledge. The code it generates is not always the best, but it works well enough for a prototype. I've felt liberated by the ability to create a working prototype of an idea that I have in my mind without deciding wether or not I want to spend weeks or months of research and programming work just to get a basic functioning prototype. That being said, the code it generates is not always the best, and there are definitely things under the hood that I don't understand. However, I was extremely intentional with designing the architecture of these projects and did my best to apply good software design principles and testing. As I expand on these projects more, I expect to spend more time manually refining the genereated output.

notes_on_ai_usage.md · rendered markdown
big_picture.md
Preview

Big Picture

I wanted to design a framework for analyzing text that could provide the reader to greater insight into a number of aspects that are often not readily apparent. For example, how phrasing may either intentionally or unintentionally create bias, how arguments are structured, whether there are logical fallacies, and more. This project is part of a larger idea that I am still expanding on, but this is a first working prototype. It breaks down input text using LLMs into the following layers: Represented Situation, Sources and Claims, Discourse, Construal, Framing & Rhetoric, and Implications and Argument. Texts that are less biased should be heavier in the categories that are more fact based, and lighter in the categories that include more rhetroic or argument strategies. This was demonstrated mostly using news articles. The prototype is in an early phase, but I encourage anyone reading this to try submitting a text, and see if you can find anything interesting in the result that you didn't notice on your first read.

Project Goals

  • Better understand how to analyze text using LLMs
  • Build a framework for examining the rhetorical structure text in an objective manner
  • Draw attention to things you might not notice on your first read

Challenges

  • Desigining a fair way to analyze rhetoric
  • Deciding on layer names and what kind of things to look for in those layers
  • Connecting the graph nodes between and across layers
project notes/big_picture.md · rendered markdown
misc_graphics_doodads.md
Preview

Graphics Doodads

I've made a few tools for creating procedurally generated graphics. I wanted to create a few designs, and give the user the ability to modify them and copy the resulting code for them to use themselves.

project notes/misc_graphics_doodads.md · rendered markdown
orbital_combat_prototype.md
Preview

Orbital Combat Prototype

This is a prototype of a multiplayer game that uses orbital physics to create tactical combat scenarios. I mostly enjoyed designing the physics engine and game mechanics, I have been stalling on the actual game design aspect, but I plan to revisit it soon. The vision is combat that doesn't rely on aim, but timing, tactics, deception and strategy. The weapons and combat style is inspired by 'Star Trek - The Next Generation'

Project Goals:

  • Build an engaging and fun spaceflight scenario that uses orbital physics as part of the game
  • Create a combat system that relies more on tactics then aim and reaction time
  • Create a cross-platform game using web technologies

Challenges:

  • Choosing when to be physics-accurate and when to take 'shortcuts'
  • Desiging and balancing weapon systems
  • Optimizing the calculations needed for gravity and missile guidance systems
  • Getting accurate 'forecast lines' for ship trajectory that weren't bouncing all over the place
project notes/orbital_combat_prototype.md · rendered markdown
sea_of_simulation.md
Preview

Sea of Simulation

The 'Sea of Simulation' is a feature of The Grid from the movie Tron: Legacy. I've been using TRON-inspired names since the premise of TRON: Legacy is a programmer creating a digital world by interacting with computer programs that can talk and reason like human characters. The process of designing this project using an AI agent resembles this concept, and the bright neon lights against an emtpy vaccum resembled the TRON world.

Project Objectives

  • Create aesthetically pleasing noise that can be generated and random or customized based on parameters
  • Inspired by corporate website backgrounds and similar abstract art
  • Create something that is entertaining enough to hit 'Randomize' a few times

Challenges

  • Surpisingly, this wasn't very hard and everything went pretty smoothly
project notes/sea_of_simulation.md · rendered markdown
snake.md
Preview

Snake

I always enjoyed playing Snake game during classes. This project was one that I hand-coded a few years ago for the first version of my personal web site, plus a little added flair from AI-Generated code.

Project Objectives

  • Create the best snake game on the internet
  • Simple gameplay, selective difficulty, retro vibe, a little flair but still minimalist.

Challenges:

  • Getting the gaussian blur effect to work on all web browsers
project notes/snake.md · rendered markdown
SUBSUNK.md
Preview

SUBSUNK

SUBSUNK is a code used by the undersea rescue command when they believe a submarine has been sunk. I wanted to design a simple game that used limited visibility and movement to create a suspensful, enemy-wave-survival experience. The submarine can only follow a few courses, and can't move very fast. Enemies are invisible until scanned by the radar beam.

Project Objectives

  • Create a suspensful game that is engaging enough to play a few rounds
  • Use limited visibility and movement to create challenge

Challenges

  • Enemy spawning and AI
  • Balancing limited visibility and movement with fairness