project file / jul 2026
Splay
A workspace for generating, reviewing, scheduling, and learning from social posts without skipping the editorial decision.
- my role
- Engineer, with the Splay team
- status
- Local-first application
- field
- Team project / publishing
- tools
- TypeScript · React · Node.js · Convex · Buffer
01 / the problem
Why this exists
A draft generator alone leaves the hardest work to a separate process: checking claims, comparing creative options, and deciding what should actually be published.
02 / the build
What I worked on
Built application and API workflows for generating posts, reviewing visual media, recording decisions, scheduling campaigns, and publishing approved work through Buffer.
Integrated durable media storage and helped harden generation and review paths for production deployment.
Kept a versioned brand kit and public-safe company context close to the draft so reviewers can understand what shaped it.
03 / engineering notes
Choices that mattered
- Long-running generation and publishing jobs are serialized to avoid conflicting changes.
- A review decision and a separate publish confirmation are required before content leaves the workspace.
04 / reflection
What I took away
Splay made editorial judgment visible in the product. It also showed how much of an AI application is ordinary reliability work: state, media URLs, retries, and clear failure paths.