At some point last year I stopped reading AI newsletters.
Not because I lost interest — the opposite. The volume got so high that anything weekly felt like archaeology by the time it arrived. A model drops on Monday. Three papers respond by Wednesday. A new eval framework ships Friday. A newsletter that aggregates these on Sunday is describing last week's internet, and the field has already moved.
I kept a browser folder instead. Open tabs I'd return to when I had time. The folder had 80-something tabs in it, all AI-related, all unread. That was my reading list.
It wasn't working. I knew it wasn't working. I just didn't have a better system.
So I built one. The thing I actually wanted was a single page that stayed current — not updated once a week, but throughout the week. Something I could open on Tuesday morning and see what dropped since Sunday. Models, papers, tools, ecosystem news — all in one place, each item with a short summary and a source link, nothing padded out to fill a newsletter format.
ai-tldr.dev (https://ai-tldr.dev/?cat=tool for tools specifically) is that thing. It auto-refreshes every 8 hours. There are category filters. Each entry is one paragraph — enough to know if it matters, short enough to not waste your time if it doesn't.
I built it for myself first. I genuinely use it to track what's shipping. The fact that other people find it useful is good, but it wasn't the original pitch. The original pitch was: I have 80 tabs open and that is not a system.
One thing I didn't expect: building the auto-update pipeline forced me to read more carefully, not less. When you're writing one-paragraph summaries that have to be accurate and specific, you can't skim. You end up with a better understanding of the thing you summarized than if you'd just read the abstract and moved on. That was a side effect I didn't plan for.
The folder of 80 tabs is gone. I close it on Monday mornings now instead of on never.