Ellery AI

Devlog #3 - Making the app more open and dynamic: I am working on making the app feel less claustrophobic, adding/updating engines like the research engine, and making the UI more professional and multipage. Not commited and running yet, but almost there.

Ellery Devlog #2 - This time wasn’t about adding new features. Instead, I spent the time refining Ellery from the ground up. I removed over 600 lines of UI code, simplified the interface, rewrote much of the copy, and cut away anything that distracted from the actual purpose of the product. The biggest lesson I’ve learned so far is that good software isn’t just about what you build, but it’s also about what you’re willing to remove. Every animation, button, label, and panel has to earn its place. Ellery feels much calmer now. The interface gets out of the way, and the focus stays where it should: understanding your data. Next up is a new demo video, and after that I’ll be taking on the first Stardance software challenge by building my own web operating system.

Ellery AI - Devlog #1 I kept watching people open a spreadsheet, stare at it, and just… give up. The story was in there. They just couldn’t see it. So I started building Ellery to fix that. Honestly, I started where everyone starts: AI that summarizes your data. But I never trusted it. The summary sounded confident, changed every time I asked, and I could never tell if it was actually true. That’s useless if you’re about to put it in front of your boss.So I went visual instead - timelines, rank boards, an insight map. They looked great. But a pretty chart still makes you do the thinking. I’d basically built a nicer haystack.The thing that actually clicked was the Discovery Engine. Before anything gets drawn, Ellery reads your data and pulls out what matters - top items, outliers, what’s out of stock - as plain findings, each one traceable back to the exact rows it came from. No model, no guessing. Same data, same answers, every time.That changed everything downstream. Now the visualizations explain themselves, and you can export the whole thing to a slide in one click.Next up: turning those findings into recommendations, then a real narrative - and pointing it at NASA telemetry to see if it holds up beyond retail.It’s early, and I’m building it in the open. Follow along if that sounds like your kind of thing.