our.directory

An alternative to AI-search

Why this exists. Posted 2026-05-26 by Tom Hamilton, founder.

The search you grew up with is gone

Type a question into Google in 2026 and what you get back is a confident paragraph of generated prose. It was written by a model that does not know whether it is right. It was ranked by a system optimised only to keep you on the results page.

Below that paragraph, if you scroll, are ten blue links. The ones that survive there are the ones that fed the paragraph. Most of them are built specifically for this algorithmic purpose. The indie blog you used to find on page one is on page four now, if it's anywhere.

The web didn't stop having good pages on it. The thing that found them for you stopped looking.

What replaced it isn't search

An AI overview is a confident summary of a search the engine ran on your behalf and decided you didn't need to see. The trade is straightforward. You get an answer faster. You don't have to click anything. In exchange, the engine takes the click economy that funded the writing in the first place. The writers notice. The pages stop being written.

There is a different way to find a page

You can ask a real person.

Before search engines, that's how the web worked. Someone who already cared about the thing you cared about kept a list of the best pages on it, and you found their list because someone you trusted linked to it.

our.directory is a modern take on one of those lists, kept by people, not AI models. It is open to anyone who wants to add to it, gated by other people who vouch for what they add. It is searchable, browsable, editable, and linkable. There is no summary written for you, no engagement metric driving what surfaces, and no secret ranking. And no ads polluting the incentive structure.

The principles

  • Every link is submitted by a named human. Their handle is on it. Edit history is public. There is no scraper feeding the index in the background.
  • Vouching is the unit of trust. When @alice vouches for a link, that is a public, attributable signal. The ranking weights vouches from people you follow more than vouches from strangers, because that's how trust actually works.
  • The ranking formula is published at /policies/ranking with the live weights. It is a linear combination of named components. It has no engagement input. We don't rank by how long you looked at a result.
  • No AI in the user-facing product. No generated summaries, no AI ranking. The one exception is admin-side: a content classifier runs on submitted text to catch things humans would want flagged. It is described further on the safety page.
  • Reader-funded. No ads, no VC, no data sales, no engagement-bait. Donations cover the bill.
  • You can leave. Account deletion is self-serve. The directory is not a trap.

If you want to help

Sign up. Submit links you actually use. Write a curator note that explains why the link is good. Vouch for other people's submissions when they're right. Report the ones that aren't. Follow curators whose taste you'd trust if they handed you a printed list.

If you write anywhere (a blog, a newsletter, anything), link to the directory.

If you can afford it, throw a few dollars in the hat. There are no ads here, and there won't be.