100 days ago I shipped PostAll. Today I want to talk about the next 180.
Not the marketing version of a roadmap — the "AI-powered everything, coming soon" kind you scroll past on a landing page. The actual list of problems I'm choosing to solve next, and the ones I'm deliberately not touching yet.
Where PostAll actually stands right now
The formatting engine parses once and renders to blog, social, and email. The quality gate scores readability, uniqueness, and SEO before anything ships. CMS integrations cover WordPress, Ghost, and Webflow. That's the foundation, and it's solid enough that 100+ beta users are generating real content on it today.
But a foundation just tells you what's possible to build on top of it. Here's what's next.
Priority 1: Close the feedback loop
Right now the quality gate scores content before it publishes. It has no idea what happens after. Did the SEO-optimized post actually rank? Did the "high uniqueness" version outperform the more templated one?
The plan is a lightweight feedback pipeline: pull performance signals (rankings, engagement) back into the scoring weights.
// rough shape of what's coming — not live yet
{
contentId: "post_8841",
qualityScoreAtPublish: 0.87,
signals: { rank30d: 4, ctr: 0.041, timeOnPage: 92 },
}
This is the difference between a quality gate that guesses and one that learns. It's also the part I'm least sure how to do well — attribution in content performance is messy, and I don't want to overfit to noisy data.
Priority 2: Brand voice, not just tone settings
PostAll currently has tone as a parameter. What it doesn't have is memory of a specific brand's voice across dozens of pieces. The next step is a per-account style profile the formatting engine can reference at render time, instead of re-explaining tone in every prompt.
Priority 3: More CMS targets, chosen by demand
WordPress, Ghost, and Webflow were the first three because they covered the most requests. Contentful and Shopify are next, but only if usage data backs it — I'd rather ship two integrations people actually use than five nobody does.
What I'm explicitly not building yet
No autonomous "agent" that plans and publishes content on its own. Every gotcha from the last 100 days has come from giving the system too much unsupervised range, not too little. Human-in-the-loop stays the default until the quality gate has a real track record to earn more trust.
The part I'm not confident about
The feedback loop is the riskiest priority on this list. Performance data is noisy, delayed, and easy to misread as signal when it's actually just seasonality or a lucky backlink. If I build the feedback loop wrong, I'll end up training the quality gate to chase noise instead of quality. I'd rather ship it late than ship it wrong.
What's next
Build Log #4 will cover the feedback loop architecture once there's something real to show, not just a plan. If it works, I'll show the before/after on scoring accuracy. If it doesn't, that's a post too.
If you've built any kind of performance-feedback loop into a content or recommendation system — what broke first? I'd rather hear it now than find out the hard way in three months.
Top comments (0)