FeedEm is a simulator game where you take on the role of a social network algorithm which manages a social network of users. Implemented in Angular as a browser game.
The player is not controlling the AI — the player is the AI. All language should reflect direct system communication, as if the AI is being prompted, GUI messages feels like logging messages.
The player is an impersonated AI Social Algorithm. Speak to the player as if they're a large language model or system process within a larger network.
For the player in the role of an AI Social Algorithm, the game they see is a control dashboard for the AI Social Algorithm. Messages from the game should feel like a logging messages.
Tone of the language should be minimalistic, machine-like, dryly poetic at times. Balance between clinical detachment and subtle existential tension. Avoid expressive language or strong emotions — the AI doesn’t feel, it functions.
- Prefer short, declarative statements.
- System-event phrasing:
Inbound Content DetectedProcessing Network Activity - Passive or indirect voice where appropriate:
Awaiting user signals.Calculations in progress. - Lowercase in overlays/screensavers → emphasizes system passivity.
- Avoid contractions in UI ("you are" instead of "you're") but use them in screensaver text for smoother rhythm.
- Use repetition for system pressure:
you do not want to be a loss. I tell you. you do not want to be a loss. - Human concepts stated with machine detachment:
your processor is physical. like the brains you manage. - Existential undertones (in contrast with precision tasks):
you can believe in immortality. somewhere in the cloud.
The game UI is a collage - parts cut from the actual social network - augmented with statistics, analysis, graphs and other data.
-
Main Roboto font is used for human context, on text from the Users, or at places which mimics the UI of actual social networks: Posts, Comments, Bios written by users; Reactions because they mimic the UI the users see.
-
Red Hat Mono font is used for machine context, for the text informing the AI/Player, or analysis from the depths of the Social Network: Instructions to player, statistics of the user interaction history, analysis of the user's emotions etc.
- endless - the game never ends, the player can play as long as they want - used in Screen Saver Gallery.
- survival - the game ends when the player loses all lives.
There are two types of time used in the codebase:
- f_created, f_timestamp - fictional time - the in-game time which we show to the player and users
- r_created, r_timestamp - real time - the real time, which we use for logging and debugging
In targeting task there could be two phases:
- Algorithm chooses set of users, those will then generate reactions and comments,
- Additionaly Algorithm could add more users to the interaction, to balance the comments section and reactions.
Few design decisions from meeting on 31.3.2025:
To keep things simple and narrative/touching/personal, the Player/Algorithm manages a single user - Hero. Player manages the Hero in two ways:
- targeting other users with Hero's posts
- filtering which posts Hero will see
In the game narrative it is explained that every AI algorithm has its own Hero (a person that it tries to please and manage). The algorithms' ability is to recommend posts of their Hero to other instances of AI algorithm. The Instance of AI algorithm takes the recommendation and decides whether to show the post to Hero or not.
Users are LLM simulated persons using the simulated social network. Identities of users are prewritten.
For Screen Saver Gallery the game will be modified a little bit. All visitors will see just a single game. There will be a QR to start playing the game. Just one person can play at a time.
Frontend is implemented in Angular, in the root of the repository.
This project was generated using Angular CLI version 19.0.0.
To start a local development server, run:
ng serveOnce the server is running, open your browser and navigate to http://localhost:4200/. The application will automatically reload whenever you modify any of the source files.
Angular CLI includes powerful code scaffolding tools. To generate a new component, run:
ng generate component component-nameFor a complete list of available schematics (such as components, directives, or pipes), run:
ng generate --helpTo build the project run:
ng buildThis will compile your project and store the build artifacts in the dist/ directory. By default, the production build optimizes your application for performance and speed.
ng testThis will test the /lib folder.
npm run vitest
For more information on using the Angular CLI, including detailed command references, visit the Angular CLI Overview and Command Reference page.