There’s far too much happening in AI to follow by hand. Papers, model releases, open-source activity, blog posts, community chatter. I built Insights so I don’t have to.

It’s private, just for me and a few people I invite. It keeps an eye on the sources I’d check anyway, notices when several of them are covering the same thing, ranks what matters, and puts together a quiet reading list instead of an endless feed.

Zig and local models do the judging

The sorting is done by my own tools, on my own hardware. Zig, my decision engine, answers the small yes/no and pick-one questions: does this item look worth a look, is this a genuinely new idea or just a product launch, does this headline actually say what happened, and what does a new reader tend to care about. Only the stories that rank highly get a proper write-up, and that comes from a larger local model.

All of it goes through my Inference Gateway. If Zig or the model is offline, the reader falls back to simpler rules, so it never sits waiting on one machine.

What I like about it

• Less noise. Overlapping coverage gets merged, so one big release is one item and not twenty.
• Ranked, not just listed. The things worth reading first come first.
• Calm by design. It’s meant to be read once a day, not refreshed all day.
• Runs on my own setup, with my own models and no per-article bill.

It isn’t public, but the idea travels well. If you’re drowning in AI news and want to compare notes on how to deal with it, get in touch.