NOVA: The "Figma for Fiction"

A collaborative publishing platform that treats writing like code.

PROBLEM

Traditional writing platforms are hostile to serious craft.

Online fiction has a discovery problem. Platforms like Wattpad and WebNovel prioritize high-frequency updates over quality, forcing writers to churn out daily chapters to survive the algorithm.

Worse, the tooling is archaic. Writers are stuck in single-player text editors, copy-pasting into CMS forms, with no version control, no real-time collaboration, and no way to branch narratives.

The Pain:

• No Collaboration: Co-authoring requires emailing Word docs.
• Black Box Algorithms: Discovery is random and exploitative.
• Zero Versioning: "What if" scenarios require duplicating entire files.

SOLUTION

A custom-built "Figma-for-writing" with real-time syncing and semantic intelligence.

NOVA (Novus Atlas) is a ground-up rewrite of the publishing stack. It brings software engineering tools to creative writing:

• Branch-Based Versioning: Git-like branching for alternate timelines and drafts.
• Real-Time Multiplayer: Google Docs-style collaboration with granular permissions.
• Semantic Discovery: A vector-based search engine that understands theme and style, not just keywords.
• Transparent Ranking: Open-source quality algorithms based on structural analysis, not just clicks.

MY ROLE

Solo Engineer & Designer.

I built the entire stack from first principles:

• Architecture: Designed the OT/CRDT hybrid sync engine.
• Frontend: Built the rich-text editor and reactive UI in Next.js.
• Backend: Implemented the vector embedding pipeline and Node.js socket services.
• DevOps: Managed the Vercel/PostgreSQL infrastructure.

THE HARD PART

Syncing state across clients while analyzing meaning in real-time.

The biggest technical hurdle was reconciling real-time collaboration with expensive semantic analysis.

1. Operational Transforms (OT) at Scale: Allowing multiple authors to edit the same paragraph required a custom OT implementation to handle conflict resolution without locking the UI.

2. The "Cold Start" Problem: How do you rank a story with zero views? I built a Structural Analysis Pipeline that scores text complexity, lexical diversity, and pacing before a single reader sees it.

3. Vector Search Costs: Running embeddings on every save is too expensive. I architected a debounced pipeline that only re-indexes changed "semantic chunks" to keep costs manageable.

ARCHITECTURE

RESULTS

• Status: Closed Beta (Wave 1).
• Performance: Sub-2s page loads via Next.js SSR and edge caching.
• Quality: 90% TypeScript 10% Rust codebase with strict linting.
• Documentation: 9,000+ word Design Philosophy published.
• Accessibility: WCAG 2.1 AA compliance (Keyboard nav, OKLCH contrast).

"NOVA represents my belief that technology should serve creators, not extract value from them."