Promo Video (After Effects / Premiere Pro)
App Screenshots
Project Philosophy:
Collective Network is an alternative social media app that strives to preserve human-centered experiential diversity in digital interaction. Collective Network embodies this in 5 pillars of operation:
1. DAO (decentralized autonomous organization) aspirations with a roadmap for crowdfunding and open source code.
2. Dashboard design that fosters digital autonomy by encouraging intentional actions with real people and omitting any algorithms or infinite feeds.
3. Core mutual aid and resource distribution framework for local collectives.
4. Ephemeral live chatrooms that foster play, surprise, and the unknown.
5. Auto deletion of most data within 90 days for digital hygiene and privacy.
GitHub - Open Source Code
Collective Network Open Source Code on GitHub
Community Spirit:
In the spirit of a collective movement, I have made the app's code open source on GitHub. Collaborators can join and pull requests for new features, fixes, etc. Someone converting this to android would be ideal for increasing the app's accessibility. Another option for users is to use the code to create their own version if they insist, or run it separately on their own servers for more privacy in their network. Ultimately, I aim to contribute tools that will facilitate human-centered digital experiences that are necessary for a new age of AI.
Initial Design and Prototyping with Figma
Collective Network Figma Design and Wiring
Wireframes into Protoyping:
The app was focused into main categories: Login, Profiles, Messages, Groups, Cyber Lounge, Events, Mutual Aid, and Confluence. The UI was developed and tested in Figma to create a core architecture, which was then moved into a proper app development phase.
HAI Collaborative Workflow Process
AI Collaboration and Claude Code:
The release of Claude Code [Opus 4.6] allowed me to exponentially accelerate my productivity. As a person with a foundation in engineering and computer science, I understand the logical frameworks required for product development, but am limited in the experience of time in specific coding languages. Being able to circumvent that and utilize AI to translate text to code, is what allowed this project to happen in 6 weeks by one person and essentially pseudo coding. In my process, I created comprehensive outcomes and Figma UI design frames that I could use to steer Claude Code [Opus 4.6] and it's generative code writing skill. I essentially had my own front and back end developer team working for me at an obscenely fast pace. The agentic component is critical in this process for understanding multi-step plans and acting on them efficiently. When those moments failed, I had to go in and manually code, even using DeepSeek at times, which was massively more labor intensive.
Claude Sonnet 4.6 Created Slide Show for a Collab Case Study
An HAI Collab Case:
Claude Code was not only critical in doing the heavy lifting of pure coding, but also amazing with developing back end security. Originally, the code was written without this in mind so had API keys hard coded and was highly vulnerable. I used Claude Code [Opus 4.6] to assist me along each level of development, shaping the architecture of the back end, and then used Claude Sonnet [4.6] for running final security audits. This process was new for me, and the slide show above (I had Claude create from our documentation) explains this along with the roles of the different Claude models.
An Example Security Brief from Claude Sonnet 4.6 in Development
The security architecture had to be established to ensure integration with the code, for example: defining and handling firestore:rules, authentication protocols, and environmental variables for API keys or secrets for all instances. This at times was nested and a little tricky to find, but resolved with tons of audits. In addition, I used an encrypted vault for called on API keys or secrets. Finally, the code has been made open source and available for community input, security review or further questioning.
Overall in this process of using Claude Code [Opus 4.6], I realized the requirement for transparency is higher than ever. The pacing and rogue nature of my work is liberating, but also could hold larger societal consequences if scaled across a whole globe of developers without accountability. Quality control, transparency, and curation are going to be massively important macro-agentic AI coding concepts we should work through together as a democratic global community.