When I started tracking AI releases for my own use, the first feature request from the first person who found the site was: "can you add search?"
The obvious answer was yes. Full-text search on 332+ AI releases feels like a must-have. Every content site has it. I even opened an Algolia account.
Then I stopped and thought about what people actually do when they land on the feed.
They don't arrive with a specific release in mind. They arrive wanting to know: "what dropped this week in the models space?" or "any interesting open-source repos lately?" or "what tools are getting traction?"
That's a browsing pattern, not a searching pattern.
The decision
Instead of full-text search, I added 5 category filters:
- models — 65 entries
- repos — 95 entries
- tools — 152 entries
- papers — 35 entries
- ecosystem — 50 entries
One click. No search bar. No query syntax to learn.
The filters at https://ai-tldr.dev/?cat=tool give you exactly the context you actually wanted when you opened the site.
Why search would have been wrong here
Full-text search optimizes for precision. When someone types "Claude 3.5" they want that exact thing.
But most people visiting an AI news feed don't have that level of specificity. They're in discovery mode. They want to see the shape of what exists, not retrieve a known document.
Category filters match that mental model better. They let you narrow scope without requiring you to already know what you're looking for.
The number that settled it
I looked at the actual query logs from my notes. The only searches I ever ran on my own feed were things like "model releases last week" or "new tools" — category-level queries, not document-level ones.
If the person building the thing only has category-level queries, the users probably do too.
What I'd add next
Eventually I want a date filter — "show me tools from the last 14 days." That's still category-adjacent. Not full-text.
Search might come eventually. But I'll need real evidence that people are looking for specific documents before I add the complexity of indexing, ranking, and query parsing.
Top comments (0)