CQI365: Discard, Refill, Repeat — Minimalism and Quality, Five Years Later

Why does decluttering always bounce back? A quality manager’s second answer.

Previously

In 2021 I wrote an essay on minimalism that split it into three layers: minimal goals, minimal processes, and minimal living. The core idea was:

Without minimal goals and processes, there can be no sustainable minimalism in material things.

Five years on, that essay remains one of the most-read posts on this site. Rereading it, I find that the core sentence was right, but I never fully explained it. A few parts were written too quickly, and there are one or two sentences I would now take back.

This episode does not repeat the method. It does three things: finish explaining that core idea, correct one cause-and-effect claim, and add a variable that only appeared in the last five years.

1. Discard, Refill, Repeat

Most of us have done this: a big weekend clean-out, three full bags thrown away. Three months later, the drawers are full again.

This is not a willpower problem. It is a location problem. Decluttering works on the output end, while clutter is a symptom of something upstream.

Quality management has an old principle: relying on final inspection will never match getting the process right at the source. Pull defective products off the end of the line, and defective products keep coming. Fix the step that creates the defect, and the problem actually shrinks.

Belongings at home work the same way. They are “produced” by your goals, habits, and buying routines. Throw things out only at the end of the line, and the upstream keeps running. Filling up again is only a matter of time.

2. Decluttering Is Only the First S of 5S

Anyone in quality management knows 5S:

Put decluttering into this table and the rebound becomes obvious: most people only do the first S.

Sorting is a one-time action. The other four S’s are the mechanisms that make its results last. In the first episode, I wrote about “standardization” under minimal processes. That was really the fourth and fifth S; I just didn’t name them.

3. Three Kinds of Waste — the First Episode Covered Only One

Lean management divides waste into three types, known by their Japanese names Muda, Mura, and Muri:

• Muda (waste): consumption that creates no value. Things you didn’t need to buy, duplicated steps, time spent looking for things.
• Mura (unevenness): a rhythm that swings between too tight and too loose. No tidying all year, then one massive year-end clean; no planning all month, then everything explodes at month’s end.
• Muri (overburden): load beyond capacity. Chasing too many goals at once, so none of them moves.

The “material minimalism” in the first episode dealt almost entirely with Muda. Looking back, what truly makes life messy is usually not Muda but the other two:

• Muri is the real problem “minimal goals” is meant to solve. Too many goals slice resources too thin, and everything ends up half-done. Half-finished things pile up: unfinished books, unused equipment, projects that were started and stalled.
• Mura explains the rhythm of the rebound. Keeping a home tidy through big clean-outs is like keeping quality high through a sprint before accreditation: the sprint ends, standards slide, and then you sprint again.

So the order of minimalism should be: first reduce Muri (prioritize goals), then smooth Mura (build a rhythm), and only then clear Muda (throw things out).

4. Taking Back One Cause and Effect

In the first episode I wrote: “Reducing material desires reduces your goals, so you won’t be confused.”

Today I see that the causality runs the wrong way. By my own three-layer model, it should be:

Reduce (or prioritize) your goals first, and material desires and belongings will shrink with them.

Desire is not the starting point; it sits downstream of goals and processes. Trying to suppress desire directly is like picking defects off the end of the line: you can hold it back for a while, but not at the source.

5. The New Variable: Execution Got Cheap

In 2021 I listed seven steps for minimal processes: inventory, selection, breakdown, scheduling, execution, review, and standardization. Even then I said the first two mattered most.

Five years later, that judgment is sharper.

AI is driving down the cost of the later steps. Breaking down tasks, scheduling, drafting, and organizing can increasingly be handed to tools, or even to agents that run them end to end. Once execution is cheap, the bottleneck moves upstream, to inventory and selection.

This creates a new kind of “discard, refill, repeat”:

• A report used to take three days, so you thought hard about whether it was worth doing.
• Now it takes thirty minutes, so “we can, so we did.”
• The result is an explosion of output: more documents, more drafts, more half-finished projects. Only this time they pile up in cloud drives, not drawers.

Digital clutter takes up no physical space, so it is harder to notice. But it still takes up attention, and it still creates confusion.

In the AI era, the heart of minimalism is no longer “do less” but “choose first.”

6. One Agent for Every Project: Minimal or Complex?

The biggest change of the past five years is the arrival of LLMs. You can now build a harness on a single LLM and have an agent handle several goals and projects at once: gathering information, breaking down tasks, drafting documents, tracking progress.

Is that minimalism, or complexity?

My answer: it moves complexity from the front stage to the back stage.

User-experience design has an old rule called Tesler’s Law, also known as the law of conservation of complexity: a system’s essential complexity cannot be removed, only moved. The only question is who carries it.

• The front stage becomes minimal: instead of switching among a task manager, calendar, notes app, and project folders, you face one entry point and one language.
• The back stage becomes complex: prompts, skills, memory, permissions, tool integrations, and error handling are all still there. They are concentrated into one system, and that system needs maintenance of its own.

Hospitals have a close parallel: central dialysate delivery in a dialysis unit. At the bedside, staff no longer mix dialysate one container at a time, so the bedside becomes minimal. But the complexity moves to the central supply room, where water quality, piping, disinfection, and monitoring all need someone in charge. A central system that nobody manages doesn’t fail one bed at a time; it fails every bed at once.

The real risk lives at the goal layer

The 2021 idea of “minimal goals” rested on an unspoken premise: human capacity is limited, so we are forced to choose. Capacity was a natural filter.

Agents remove that filter. A harness can push ten projects at once, so naturally, you take on ten projects. This is the Jevons paradox from economics: greater efficiency leads to greater consumption. In the language used earlier, agents reduce Muda but can massively amplify Muri.

And the overload changes location. It used to be “I can’t finish it all.” Now it is “I can’t review it all.” Ten parallel streams of output each need someone to read them, judge them, and take responsibility. People shift from doers to reviewers, and attention becomes the new bottleneck.

It also has a genuinely minimal side

From a quality perspective, a harness turns processes into executable standard work: skills are SOPs, memory is the handoff record, permissions are the scope of delegated authority. This is 5S’s Standardize and Sustain, and PDCA’s standardization step, made institutional. The first episode stopped at “standardize”; agents let that step actually run on its own for the first time.

So my conclusion is:

The interface is minimal and the system is complex. Whether your goals are minimal depends on whether you still choose to do less after your capacity grows.

Two questions for self-checking:

1. Since adopting agents, are you running more projects at once, or fewer?
2. Do you have more to review each day, or less?

If the answer to both is “more,” what you have is not minimalism. It is a more efficient form of overload.

7. Tool Minimalism: An Honest Self-Review

The process in the first episode lived in OmniFocus. Look at this blog’s tag cloud, and other tools from the past five years appear too: Akiflow, NotePlan, Logseq, Obsidian, DEVONthink… In the end, my task manager settled on Todoist.

Someone who writes about minimalism switching tools this many times deserves a review in its own right.

Migrating tools has real costs: moving data, rebuilding workflows, re-forming habits, and the mess of running two systems in parallel during the transition. Sometimes switching tools is optimization. Sometimes it is a way to avoid “doing the work” by “organizing the system,” which gives you the illusion of progress.

So I now ask myself three questions before switching:

1. Did the workflow change, or did my mood change? Tools need to change only when the work itself has changed.
2. Exactly which step is the old tool blocking? If you can’t name the step, the tool is usually not the problem.
3. Have I counted the migration cost? Including the transition mess, not just the time to move data.

Only if I can answer all three do I switch.

With agents, this takes on another layer of meaning. When an agent sits on top of your task manager, the front-end tool itself matters less. The task app shifts from “the system I operate every day” to “a data layer the agent reads and writes.” Which app has the nicer interface matters less; what matters is where the data lives, whether the format is stable, and whether it can be read and written reliably.

In other words, the tool wars end on the front stage and are only beginning on the back stage.

8. Taking Back One Sentence

In the conclusion of the first episode I wrote: “Truly capable people are efficient at work, and even the happiness of their time off is efficient!”

I would like to take that back.

It contradicts the “mindfulness” I cited in the same essay: when you play, you play, and you shouldn’t be measuring play for efficiency. Putting leisure under optimization is exactly why so many people find efficiency culture exhausting: even relaxing becomes another to-do list.

My version, five years later:

The point of minimal processes is not to make every minute productive, but to give the empty space back to yourself.

Efficiency is a means. If the time you save just fills up again, that is only another round of “discard, refill, repeat.”

Closing

Taken together:

Minimalism was never a single big clean-out. It is a quality system that keeps running. That is what CQI — continuous quality improvement — was meant to say from the start: not getting it right once, but fixing the upstream, every day.