I am not a developer. I want to be clear about that upfront, because what I’m about to describe sounds like something a developer would do, and I think the fact that I’m not one is the whole point.
I sat down one weekend with Claude – Anthropic’s AI – and started talking. Not coding. Talking. I described what I wanted. Not in technical terms, but in GTD terms. I want a system that holds my projects and next actions. I want it to tell me what’s due today. I want it to notice when something is stalled. I want it to stop me from avoiding the hard stuff.
Claude listened. It asked questions. It pushed back on things that wouldn’t work. And then it started building. I’d describe a behavior, and within minutes there was working code. I’d test it, say what was wrong, and it would fix it. We went back and forth like that for hours, and by the end of the weekend, I had something that actually worked. Not a prototype. Not a mockup. A real tool, running on my own server, that I use every day.
A finance guy building production software in a weekend. That sentence would have been absurd two years ago. It isn’t anymore.
But the tool isn’t the interesting part. The interesting part is what happens inside it.
I have an agent that lives inside my GTD system. It reads the same files I do – the same project lists, the same next actions, the same waiting-fors. Every morning, before I’ve even poured my coffee, it’s already done the scan. It sends me a briefing: here’s what’s due today, here’s what’s overdue, here’s the project you haven’t touched in two weeks, here’s the one where you’re “waiting for” a reply that never came.
That last one is the killer feature. Not the due dates – any calendar can do that. The noticing. The agent surfaces the things I would skip past in a manual review. The project where the next action is vague and I’ve been avoiding clarifying it. The waiting-for that I told myself I’d follow up on “soon” three weeks ago. The context I keep assigning work to but never actually get to.
It does the noticing so I don’t have to.
End of day, another scan. Here’s what you completed today. Here’s what carried over. Here’s what showed up in your inbox that hasn’t been processed. It’s not a nag. It’s not a guilt trip. It’s more like a colleague who keeps good notes and isn’t afraid to mention the thing you’d rather not look at.
And here’s what changed about the weekly review: it isn’t a grind anymore. By the time I sit down to review, most of the noticing has already happened. The stale projects have already been flagged. The vague next actions have already been called out. I’m not starting from scratch every Sunday, dreading the pile. I’m confirming what I already know and making a few adjustments. Twenty minutes instead of an hour. And I actually do it, which is more than I could say for the last three years of app-based systems.
The system never gets heavy. That’s the thing I couldn’t achieve with paper or with any app I tried. Something is always working it. By the time I look at my dashboard, the tending has already happened. There’s no backlog of 47 overdue items staring at me with red badges. There’s just today – what’s on it, what’s coming, what needs a decision.
I pull up the dashboard on my phone while I’m waiting in line. On my tablet on the couch. On my laptop at work. The capture problem is gone – I can add something from anywhere, and it’s immediately part of the system. The notebook’s constraint was geography. This has none.
But I want to be honest about what really shifted, because it’s not the technology. It’s the relationship to the technology.
I used to be a consumer of tools. Someone else designed the app. Someone else decided what the review screen looked like, what “overdue” meant, how projects and tasks related to each other. I fit my thinking into their structure, and when the fit was bad, I blamed myself for not using it right. That’s what everyone does. The app doesn’t work, so we assume we’re the broken part.
Now I shape the tool. When something isn’t working, I don’t search the settings menu. I describe the problem, and the tool changes. Last week the morning briefing was too long – it was listing every active project regardless of whether anything had changed. I said “only show me projects where something actually shifted or where something is overdue.” Ten minutes later, it was fixed. Not because I wrote the code. Because I described the behavior and the partner implemented it.
That’s a different relationship entirely.
David Allen’s methodology didn’t change. I still process inputs. I still define outcomes. I still clarify next actions. I still review. The five stages of mastering workflow are the same ones I learned from the book fifteen years ago. The skill is the same skill. But the implementation – the actual living experience of it, day to day – is unrecognizable from where I started.
The notebook was good. It was the right tool for a long time. But it was a tool that waited for me. If I didn’t show up, nothing happened. The lists sat there, getting stale, until I opened the cover and did the work. Everything depended on my discipline, my energy, my willingness to face the hard stuff on a Sunday morning when I’d rather not.
The partner doesn’t wait. It works the system whether I’m there or not. When I show up, it’s ready. When I don’t, it keeps going. The weight that crushed every system I ever used – that slow accumulation of guilt and avoidance – doesn’t build up, because something is always relieving the pressure.
The medium changed. The implementation changed. But the skill – the skill is mine. It always was. And now, for the first time, the system works with me instead of waiting for me to show up.
Mind like water. For real this time.
Want to see what it looks like? Try the interactive demo – no signup, no login, fully interactive with sample projects.