Silknode

Silknode

founding engineer for next-gen blogging and curation platform

Remix // Postgres // Drizzle ORM // Fly.io // S3 // Redis
overview
challenge
approach
outcome

Overview

Silknode is building a social platform that reimagines content curation. Users collect media blocks (images, videos, links, text) into "webs" that can be nested and shared. The platform encourages posting content as an act of community rather than ownership—users contribute to a shared body of content by posting. I joined as the founding engineer and first hire just before their seed round, responsible for building the product from scratch and making all technical decisions.

Challenge

The product philosophy required unique information architecture: content blocks that persist independently of their creators, automatic content deduplication, and a unique approach to the concept of following. Proper implementation of this IA at the database and application levels was crucial to supporting the product vision.

Beyond architecture, the project demanded solving several complex technical challenges on a tight timeline:

Media processing at scale

Link previews required screenshot generation via headless browsing, which could take several seconds and needed careful memory management.

Sophisticated UI interactions

A responsive media grid supporting masonry, justified, and standard grid layouts with dynamic gap sizing—all needing to feel instant and joyful.

Performance under growth

When alpha users posted 100k+ blocks faster than anticipated, optimization of feeds became critical, requiring rapid implementation of virtualization and infinite scroll.

The product was built for aesthetic-focused end users, so it needed to feel polished and performant from day one.

Approach

Greenfield architecture

Evaluated and scrapped existing codebase within first week, starting fresh with a carefully researched tech stack optimized for low cost, ease of use, and ability to scale to thousands of users with minimal infrastructural overhead.

Database design

Modeled complex relationships to support unique IA while keeping queries performant.

Screenshot pipeline

Built asynchronous worker using BullMQ and Puppeteer to generate link previews, with server-sent events to update UI once screenshots were ready, managing headless browser memory constraints.

Custom grid component

Implemented responsive media grid that calculates absolute positioning for each block based on view type, enabling instant view switching without re-rendering the entire grid.

Rapid optimization

Implemented virtualization and infinite scrolling for feeds in approximately one day when user growth outpaced expectations.

Self-directed execution

Decomposed six-month roadmap into daily tasks, maintained long-term progress tracking, and made unilateral technical decisions working directly with founders.

Outcome

Built the product from concept to working alpha in approximately six months, delivering core features including image and link posting, web creation and editing, following mechanisms, and dynamic feeds. The platform successfully launched to alpha testers and continues to operate today, supporting 1,800 users who have posted over 350,000 blocks.

The technical foundations proved robust and extensible—other developers have continued building on the codebase, adding new features while the core architecture scales to handle ongoing user growth. The product demonstrates that thoughtful early technical decisions can create lasting value even as teams evolve.

Study: Working Families Party
|
Study: TITLES