Developer Reflects on 100 Days Building AI Content Platform PostAll in Public

A solo developer has published a 100-day retrospective on building PostAll, an AI-powered content platform that grew from a basic OpenAI API script into a full system with CMS integrations for WordPress, Ghost, and Webflow. The builder's original assumption — that AI content generation was the hard part — proved wrong within two weeks, as structuring and rendering output consistently across formats turned out to be the real challenge. A key architectural win was a 'parse once, render many' model, which separated content generation from platform-specific rendering and simplified later CMS integrations. The project's biggest misstep was an inadequate uniqueness detection system using Jaccard shingling, which failed to catch semantically similar paraphrased content and was only exposed by a beta user complaint. The developer ultimately combined Jaccard with sentence embeddings to address both literal and semantic duplication, and acknowledged that the architecture also failed to scale gracefully beyond a small user base.