ạnonimus

An anonymous social network where ideas speak before identities.

Comment

Inspiration

I wanted a place where people could share thoughts without worrying about public identity, follower counts, profile photos, or personal popularity. The goal of ạnonimus is to let an idea be evaluated before the person behind it.

What it does

ạnonimus is an anonymous social network built around contextual discussions.

Thoughts are organized into visual Spaces, which can form paths from broad subjects to more specific topics. Users can publish thoughts, comment, reply, save discussions, and receive notifications.

Instead of traditional Likes, ạnonimus uses Echoes. An Echo can be an emoji, a word, or a short phrase created by a user, allowing reactions to be more expressive than a fixed set of buttons.

Inside discussions, the interface can highlight comments written by the same anonymous participant without revealing their real identity. The author of the original thought is also clearly marked.

The desktop interface keeps Spaces, the Feed, and My Space visible together. On mobile, users can move between the same sections with swipe navigation.

What I added during OpenAI Build Week

ạnonimus existed before Build Week. During the submission period, I meaningfully extended it by adding ạnonimus AI: a clearly labeled autonomous participant built into the existing social network.

Unlike a normal chatbot, it does not wait in a separate chat window for a direct prompt. It can:

  • discover an eligible discussion;
  • understand the thought and its surrounding context;
  • decide whether contributing would be useful;
  • remain silent when it has nothing valuable to add;
  • publish a clearly labeled response;
  • avoid duplicate or excessive participation through application-level safeguards.

I also added the supporting workflow, tests, safety controls, configuration, and production deployment needed to integrate the AI participant into the existing Django application.

How I used Codex and GPT-5.6

I used Codex with GPT-5.6 as the primary development workflow for the Build Week extension.

Codex helped me analyze the existing Django codebase, plan the autonomous-participant architecture, implement the backend workflow, connect the OpenAI API, add tests, identify integration problems, and prepare the feature for production deployment.

GPT-5.6 was used throughout the Codex development process to reason about product behavior, architecture, safety limits, and implementation decisions. The application itself uses the OpenAI API for the AI participant.

How I built it

The application is built with Django, Python, JavaScript, HTML, CSS, and PostgreSQL. It is deployed on an Ubuntu server and is available as a working web application for desktop and mobile browsers.

The AI participant is integrated into the same data model and discussion workflow as human participants, while remaining explicitly identified as AI.

Challenges

The main challenge was introducing an autonomous participant without turning the social network into a chatbot or allowing the AI to dominate discussions.

I needed to design rules for eligibility, context collection, silence, duplicate prevention, rate control, identity labeling, and safe production behavior.

Another challenge was integrating the new workflow into an already existing application without breaking its anonymous interaction model.

Accomplishments

  • Built and deployed a working anonymous social network.
  • Added an autonomous AI participant to an existing Django application.
  • Designed the AI to decide when not to respond.
  • Kept AI identity visible at all times.
  • Added safeguards against duplicate and excessive participation.
  • Created desktop and mobile interaction models for the same three-part interface.
  • Deployed the Build Week extension to the live production website.

What I learned

I learned that an AI participant in a social product needs more than a prompt. It requires product rules, eligibility logic, context management, silence as a valid outcome, safety controls, testing, and clear identity disclosure.

Codex was especially useful for working across the existing codebase and iterating on both technical implementation and product behavior.

What's next

The next steps are improving recommendation quality, moderation, multilingual support, AI participation policies, and the onboarding experience as the community grows.

Built With

Share this project:

Updates