Today's work didn't involve a visible UI feature. It was infrastructure, the recommendation algorithm that decides what content users see the moment they open the app. And I kept it intentionally simple.

The Core Idea

During onboarding, users select interest categories like Science, Development, Design, Finance, and so on. Those preferences are stored on their profile and become the single source of truth for everything they see on the homepage feed.

The 70/30 Rule

When the homepage loads, posts are fetched, scored against the user's interest list, and split into two buckets before rendering the feed.

• 70% of the feed = posts that match the user's selected interest categories
• 30% of the feed = exploration content from categories outside their interests
• Within each bucket, posts are sorted by recency so the feed always feels fresh

No machine learning. No behavioural tracking. Just your preferences as the source of truth.

Why Keep It This Simple?

1. Complex algorithms need data

A weight-based or collaborative filtering system needs meaningful user behaviour signals to produce good results. At this stage, I don't have that data. A sophisticated algorithm running on thin data doesn't produce smart recommendations, it produces confidently wrong ones.

2. Premature complexity kills momentum

If I tried to build something clever this early, I'd spend two weeks architecting edge cases and ship nothing. The right order of operations at this stage is simple: ship, learn, then improve. Complexity has to be earned.

What I Also Shipped Today

The homepage UI now renders a proper feed immediately on login. No search required, no blank state staring at you. You open Conjure and you start reading. That friction removal matters more than people realise, the best feature is the one that gets out of the way.

What Comes Next

The 70/30 split is a starting ratio, not a final decision. I'll tune it once real usage data shows whether users want more familiarity or more discovery. A learning mechanism that factors in reading history is on the roadmap, but it earns its complexity only after the simple version has been validated by real users on real content.

The algorithm is only as good as the content that feeds it. The more blogs get written on Conjure, the better the recommendations become.

If you have been thinking about writing on Conjure, now is a genuinely good time. Every post makes the platform smarter for every reader. Come write something at conjure.blog.