Launch note · 7 min read

Why I built a LinkedIn operating system—not another post generator.

A good post is an output. A good campaign is a system that remembers, measures, and chooses what comes next.

The blank-box problem

Most AI content workflows begin from zero every time. They ask for a topic, draft a post, and end. The research, engagement context, performance evidence, and unfinished work live somewhere else—or disappear.

The campaign became the product

I started with a specific goal: build an organic system for 10,000 followers and 10,000 connections. The implementation grew into eight specialist skills coordinated by one stateful operator.

State outside the skill

The most important architectural decision was to separate versioned instructions from mutable campaign learning. The plugin can improve while campaign configuration, logs, analytics, experiments, and verified actions remain portable and inspectable.

Public by design

Version 1.0 removes the single-profile assumptions. Identity, profile URL, timezone, niche, baselines, and goals are configurable. The same package now supports Claude Code and Codex, and the whole project is available under the MIT license.

Build with it

Install it, run one workflow, and share the artifact. The next phase is a library of real campaign examples, adapters, and community-built skills.