Memex

Signal in a complex world

You typed what you thought. Underneath, a hypergraph formed — one note connecting to multiple people, concepts, and prior notes simultaneously, with paths forming across weeks of thought. Then someone else's post from the network folded right into the same graph.

You didn't organize any of it. You didn't tag, categorize, or file. You thought out loud and the structure emerged. That's the point — complexity is the default, but your view of it doesn't have to be.

Beyond Dunbar's number

You can meaningfully follow about 150 people. But the world's knowledge doesn't fit in 150 heads. Feeds and forums scale by showing you more — more posts, more threads, more noise. That doesn't work. More content with the same attention means less signal per item.

A hypergraph scales differently. You don't follow people — you follow the knowledge. When someone you've never met posts something that connects to a concept in your graph, the structure surfaces it. When 200 people discuss a topic across 50 threads, the graph collapses their contributions into the concepts that matter to your existing understanding.

This is many-to-many without a feed. Contributions from people you've never spoken to fold into your graph when they touch the same concepts. The more your graph grows, the more signal the network can deliver — because it knows what you already understand.

Effective attention

The internet gives you everything. That's the problem. Your attention is the scarce resource, not information. Every tool you use should make your attention more effective, not compete for it.

Memex is a framework for this. Locally, your knowledge graph grows with every conversation — connecting new ideas to everything you already know. On the network, trusted peers and relevant topics surface automatically. What's new to you and connected to what you care about rises. Everything else stays quiet.

You follow keys, not accounts. Reputation accrues per domain — someone insightful about distributed systems isn't automatically worth listening to about nutrition. The structure is rich. The interface is simple. A terminal and natural language.