WorldSieve
Live URL: worldsieve.com
Role: Founder & Architect (Shoestring Startup / AI Dark Factory)
WorldSieve is an ambitious procedural world-generation and geospatial simulation platform built using an autonomous AI Dark Factory paradigm. It marries deep scientific computing with an ultra-responsive, interactive canvas web interface.
The AI Dark Factory Paradigm & Agentic Workflows
WorldSieve was engineered as a shoestring startup by establishing a hardened dark factory development pattern:
- Local Frontier AI Workflows: Utilizing Google Gemini frontier models alongside local Ollama instances, integrated through the Model Context Protocol (MCP) and custom RAG indexers.
- Strict Quality Gates: Autonomous agents operate against codified architectural manifests, producing PRs that must satisfy exhaustive unit test suites (
pytest,vitest), TypeScript type checking, and zero-lint policies before merging. - Shared Contextual Memory: A centralized codebase vector and AST index provides continuous architectural intent to agents, enabling complex multi-file refactors without drift.
- Continuous Velocity: What traditionally required a team of 8 full-stack engineers is executed by a single founder orchestrating autonomous agent swarms.
Technical Architecture & Stack
Frontend
- Core: React 19, TypeScript, Vite, React Router DOM, Styled-Components.
- Rendering & Geospatial: Paper.js for high-performance 2D vector canvas rendering, d3-geo for complex spherical and planar map projections.
- UI & Observability: Modular floating toolbars, interactive fog of war, Lucide icons, Sentry error monitoring, PostHog product analytics.
- Testing: Vitest, React Testing Library.
Backend & Distributed Computing
- API Runtime: FastAPI, Uvicorn, Python 3.12, Pydantic, HTTPX.
- Database & Spatial Engine: SQLAlchemy 2.0, GeoAlchemy2, PyMySQL, spatial indexing for complex geometry.
- Scientific Computing: NumPy, SciPy, PyProj, GeoJSON, OpenSimplex noise, and Pillow for terrain generation and tectonic math calculations.
- Async Task Queue: Redis + ARQ background workers executing heavy Voronoi polygon cell generation and hydraulic erosion algorithms.
- Security & Infrastructure:
fastapi-users, Cryptography, SlowAPI rate-limiting, AWS ECS containerized deployment via Bitbucket Pipelines.