A lot happens over the course of a workday that needs attention after the work itself is done. A meeting produces a follow-up. An email changes the next step in a project. I promise someone a resource, discover something worth keeping, or finish a task that makes another task possible. All of that needs to go somewhere. And, at the appropriate time, it needs to come back. That’s the job I’m giving my “personal agent.”
I’ve been building a system that functions as an executive admin: helping me file, organize, and manage the work I do, then reminding me of what needs to happen next. I work with it through Codex, and its home is HQ, a system created by R.J. Nestor. His approach connects everyday actions to the projects and responsibilities they serve. His work also helped me see the possibilities of bringing AI directly into my files and workflows. R.J.’s work was an important starting point.
Another substantial influence is Nick Milo’s AIOS, which provides a way to build an AI assistant around your own knowledge, preferences, and working methods. My Daily Brief came over from my use of AIOS. I also brought along task and project management practices developed during my years of using Tana. What I have now combines those influences with the very particular ways I work. R.J. and Nick helped open up the world of “what can be done.” Using their ideas gave me enough of a starting point to begin asking what I wanted done.
The personal part matters here. My system needs to help me keep track of library support work, presentations, ongoing projects, reading, and the commitments that accumulate in conversations. Someone else’s might need to handle an entirely different set of responsibilities. Even when we need similar things, we may want to see them differently or be reminded at different times.
HQ gives me a foundation I can change as I discover those preferences. And this is where things get interesting: the system helps build itself. I can describe something I need, work through how it should behave, and have the agent build or adapt the pieces. Then I use it in an actual workday and find out what needs adjusting. Those adjustments become part of the system.
A practical note if you try this: make sure you have plenty of AI credits available for that first week or two. There can be a lot of building, testing, and “now that I’ve used it, could we change this?” involved. Figuring out what you want is part of the work.
The most visible piece of my setup is the Action View dashboard, created entirely by Codex—the app I downloaded to my computer to build out my system. I didn’t need to do any coding or web design myself; the AI agent essentially built itself, with me describing what I wanted and asking for changes. The dashboard gives me a place to look at my day, see tasks alongside calendar commitments, review projects, and update the work as I go. I can talk to the agent when I need help, and I can use the dashboard when I want to see what’s happening.
The Daily Brief provides context around that view. It brings together upcoming commitments, recent activity, loose ends, and suggested next steps. It also makes room for reading and ideas. “What You Might Be Forgetting” is a particularly useful thing to have someone—or something—help you consider.
Underneath both are the task and project practices I brought from Tana. Individual tasks need a connection to the larger work. Projects need review. Something I intend to do eventually needs to be distinguishable from something that needs my attention today. That’s how the pieces help me get the right stuff done as I go through my days. I still decide what matters. The agent helps preserve those decisions, organize the work around them, and bring them back when I need them.
When a day gets interrupted, I have somewhere to return to. When I finish a conversation, I have help catching the commitments. And when I sit down to work again, I don’t have to reconstruct everything from memory. That’s the kind of personal agent that can be really useful in any library or really, any office situation.