Building in public — Kyle Covan

Project 01

Personal AI OS Nate Herk style

An AI operating system that runs quietly in the background like a digital assistant, flagging opportunities and spotting bottlenecks before they become real problems.

What the log covers

The friction that started it — the specific thing in my own week that kept slipping.
How it works today — the tools it connects, what it watches, what it surfaces to me and when.
What it caught — concrete examples of an opportunity flagged or a bottleneck spotted early.
What broke — the versions that failed and what I changed.
Where it goes next — the next capability I'm adding.

Build log

July 28, 2026

Too many ideas, and no way to rank them

I tend to have a lot of ideas. I'd start a project, then another idea would pull me sideways, and the first one would sit there unfinished. What I needed wasn't more ideas. I needed somewhere to put them, and something to tell me which ones were worth my time.

So it keeps track of all of them and gives them a ranking. That shows me which ones to focus on first, which ones compound, and which ones aren't worth putting time or energy into. The point is to keep me structured enough to focus deliberately instead of getting sidetracked by whatever idea came last.

Today it watches my calendar, my Gmail, and my Obsidian vault, which is where my notes and random thoughts end up. It's connected to OBS, so it records for me. And it takes a rough idea and fleshes it out into structured steps in a roadmap, so I can see what's next and how to go about it.

Here is the thing it catches that I couldn't. I'll have the same idea on different days, worded differently each time, and not notice. It sees that they're the same thing and puts them together.

Next is a pipeline for making YouTube videos and social content in a streamlined way, so most of that work comes off my plate.

Project 02

Second Brain Andrej Karpathy style

An AI-powered knowledge base that centralizes internal documentation and unstructured data so hard-won knowledge is never lost.

What the log will cover

The problem — what I was losing track of, and where it was scattered.
How it's structured — ingestion, indexing, and how I actually retrieve things.
Retrieval in practice — a real question asked and the answer it returned.
What I'd do differently — lessons from the first architecture.
Where it goes next — the next source or capability I'm wiring in.

Build log — first entry coming soon.