# Three loops: how I run my week

> A monthly vision pass, a weekly build pass and a Friday audit. The filters I use to decide what to build, and why I commissioned a control-tower system.

8 Oct 2026, by Ali Asif.

At one point I typed this into a chat window: "What's built, bro? I'm confused, there's so much text, I don't know what the next step is."

I'd built so many tools and skills that I couldn't remember half of them. They sat idle because I couldn't recall making them. The real cost of running a company on AI isn't compute. It's the day your own system grows past what your head can hold.

This matters because I'm a good starter and a poor finisher, and I think by building. I generate options, react to them, and the building becomes the deciding. It's a useful engine that also produces clutter.

So I stopped trying to fix my memory and built structure around it.

## The three loops

**Vision, monthly.** Once a month I look at the whole board and ask what I'm actually trying to do. Not what I'm doing. What I want to be true in a year. This is where I decide which ventures stay alive.

**Build, weekly.** Weekly is when I make things, and I pick the targets before I sit down, because otherwise I'll build whatever is most fun.

**Audit, Friday.** I check what I said I'd do against what happened. Nothing I audit is allowed to sit as "in progress" forever. It either moved, or it goes on the kill list.

## The filters

Before I build anything, it has to survive two checks.

The first is a two-by-two. One axis runs from easy to hard. The other runs from compounds to doesn't compound. Easy things that compound get done now. Hard things that don't compound are the traps. They're the projects where the effort is large and nothing carries forward afterwards.

The second is a 90-day payoff filter. If I can't name what this changes within 90 days, it waits. I made this rule up for myself after I'd started several things I never finished.

Every quarter I run a kill list. I write down what I'm stopping, and I'm allowed to mourn it for an afternoon. Ambition without a kill switch becomes clutter.

An idea that survives has to pass one more gate before it runs on a schedule: a trigger, a cadence, a heartbeat that tells me it's alive, somewhere I can see it, and a condition for killing it. If I can't fill all five in, it stays a note.

## The control tower

The system I commissioned is a dashboard over everything. It reads the files themselves and works out what's current, what's stale and what I dropped, using the date each file last changed.

The design came from a failure. Every notes system I tried died the same way. A status field that a human updates by hand goes stale, and then the board is a lie. So the signal has to update itself, because I won't.

I also learned that modules which produce something stay alive. Modules that only record state die. In my own dashboard, a web page I built to show the pipeline had zero opens in 90 days, though I edited files in that repository every week. The data was fine. The page lived somewhere I never went, so I moved the output into the places I already look.

I commissioned all of this because I know how I work. I forget what I build. That's a design input, not something to be ashamed of.

## What it hasn't fixed

I still start things I shouldn't. The filters slow me down, they don't stop me. I have scoped ideas that the 90-day test would have killed in two minutes, because I wanted to see how they looked. The gate works when I run it, and I'm the one who has to run it.

The Friday audit also has a soft spot. It's easy to be generous with myself.

If you recognise yourself in this, don't begin with an app. Begin with the kill list. Write down three things you'd stop, and see how you feel when they're gone.

Canonical: https://heyaliasif.com/writing/2026-10-13-moves-three-loops-how-i-run-my-week