# What do you delete before you automate?

Momo Maximilian Feichtinger · Education · planted 2026-09-02, last grown 2026-09-03 · 5 min · https://zukunftbilden.eu/en/denken/loeschen-vor-automatisieren

**When a tool or a ritual at school feels too heavy, too slow or too expensive, we run the same five steps, in the same order, every time. Automation comes last. Bringing something back is cheap, carrying it along is expensive, and that is exactly why deleting comes so early. The same order applies before an ki-department takes over a process.**

![Five steps in a row: question mark, eraser, scissors, hourglass, gear](https://zukunftbilden.eu/assets/denken/loeschen-vor-automatisieren.webp)

*The mental model: question, delete, simplify, speed up, automate. The gear comes last.*

## Five steps in a fixed order

The first step asks about the requirement itself, in the sense of first-principles: an idea proves itself against what is fundamentally true for children and their education. Who decided it has to be this way, and what happens if it goes away. Every requirement needs a name and a reason, or it is only a rumour.

### The improvement algorithm

1. Question the requirement. Who set it, and what happens if it goes away.
2. Delete. Cut so much that you have to bring something back later.
3. Simplify what is left.
4. Accelerate the simplified process.
5. Automate last. Only once a process has been questioned, stripped, simplified and accelerated may technology take it over.

> Whoever starts at step 5 automates things that should not exist at all.
>
> Paideia core handbook, section 20.5

Existing things have to justify themselves again too. Every tool and every ritual needs a reason to continue, and what no longer has one is cut, however familiar it is. Effort, habit and pride do not count as a reason. To make that happen, improvement has fixed slots: 30 minutes of learning intake every month, 90 minutes of review against the axioms every quarter, the last 15 of them as the cutting round, and once a year a red-team of two to three hours. Every decision lands in four fields: observation, decision, owner, review date.

## Where the two ideas come from

Kaizen means continuous and never ending improvement in small steps. [Masaaki Imai](https://en.wikipedia.org/wiki/Masaaki_Imai) made the term known with his book from 1986, and the root lies in the [Toyota production system](https://www.lean.org/lexicon-terms/kaizen/). Thinking from the foundations goes back to [Aristotle and his Metaphysics](https://plato.stanford.edu/entries/aristotle-metaphysics/). Both hold constitutional rank in the paideia core handbook, and the improvement algorithm is their practical form. The school wrote this up at length in [Kaizen and first principles at school](https://freieschulesalzburg.at/blog/die-schule-die-nie-fertig-ist).

![Drawing of a small cymbal with a mallet on a wooden table, next to a single board computer with a blue wire and a hand drawn construction plan.](https://zukunftbilden.eu/assets/denken/loeschen-vor-automatisieren-2.webp)

*The improvement algorithm played through on a single object, from questioning to automating.*

## The cymbal, the timer and the striking robot

One example from daily operations shows the order on a single object. A cymbal gives the acoustic signal for the morning circle, the change of blocks and the reflection at the end of the day.

### One signal, step by step

- **Until August 2026** A mentor strikes the cymbal by hand. The ritual is proven before anyone thinks about technology.
- **12 August 2026** The decision to make the signal reliable. The obvious answer, an automation on a phone, fails the test: battery optimisation, confirmation dialogs and restarts make it unreliable.
- **First week** A phone takes over anyway, as the stopgap, so the day runs while the durable version is built.
- **After that** A Raspberry Pi with cron, the standard scheduler from Unix systems, becomes the timer. After a power cut it starts again by itself.
- **As a build project** The children build a robot cymbal that strikes the real cymbal mechanically.

The order is the point. Lived by hand first, then simplified, then automated, and in the end the automation becomes a project of the children themselves.

I had to work this order out for myself first. In 2018 I had a scheduling tool of my own programmed and handed it straight to beta testers, even though the basic function was not yet reliable: entries disappeared for no obvious reason, and I had to warn the testers about it myself in a video. I had automated before the process was robust, exactly the mistake this order is meant to prevent today.

## The questions before any automation

Every planned automation faces one test: does it measurably move learning, or does it save a mentor time? If the honest answer is that it would just be handy, it falls under the rule to cut. The same questions work at any desk.

### Guiding questions

- Who really needs the result?
- What happens if nobody does it for a month?
- What is left if you cut the task down to one sentence?
- Does it move something measurable, or would it just be handy?

*An interactive tool sits here on the page: https://zukunftbilden.eu/en/denken/loeschen-vor-automatisieren*

## What this means for a company

A company has the same chain, only with different names. First the requirement, then deleting, simplifying, accelerating. Only what survives these four steps earns an agent at the end. Anything else would be technology built around something that should not exist. In practice: the delete list comes before the first prompt, and the first agent is still [the guardian of the brand](https://zukunftbilden.eu/en/denken/waechter-der-marke). How a department grows out of that stands on [AI implementation for companies](https://zukunftbilden.eu/en/ki-umsetzung).

The second half of the chapter matters just as much: question before you copy. The school learns from others and adopts nothing blindly, no cash prizes for children, no screen surveillance, no status contest over percentile ranks. The comparison with a well known school in Texas stands in [Alpha School and Freie Schule Salzburg](https://freieschulesalzburg.at/blog/lern-eingang-alpha-school).

> This handbook is itself subject to kaizen. Anyone who finds it unchanged in ten years has used it wrongly.
>
> Paideia core handbook, section 20.6

### On record

- Five steps in a fixed order: question the requirement, delete, simplify, accelerate, automate
- Deleting counts as more effective than automating
- The cymbal was struck by hand first, then made reliable, then turned into a build project for the children
- Fixed slots: 30 minutes of learning intake a month, 90 minutes of axiom review a quarter, one red team a year
- Existing tools and rituals have to reprove their reason to exist

Source: Paideia core handbook, chapter 20 (The school that is never finished), sections 20.1 to 20.6. Section 20.5 added on 13 August 2026. Cymbal decision of 12 August 2026.

Download: [Department worksheet](https://zukunftbilden.eu/assets/denken/downloads/loeschliste-en-2026-09-03.pdf) · PDF · A4, 1 pages, 41 KB

### Sources and links

- [Kaizen and first principles at school: why we are never finished](https://freieschulesalzburg.at/blog/die-schule-die-nie-fertig-ist) · The school’s long form piece on the same chapter. In German.
- [Alpha School vs. Freie Schule Salzburg: AI, mastery, relationship](https://freieschulesalzburg.at/blog/lern-eingang-alpha-school) · What was deliberately left out, and why. In German.
- [Kaizen in the Lean Lexicon](https://www.lean.org/lexicon-terms/kaizen/)
- [Masaaki Imai](https://en.wikipedia.org/wiki/Masaaki_Imai)
- [Aristotle, Metaphysics (Stanford Encyclopedia of Philosophy)](https://plato.stanford.edu/entries/aristotle-metaphysics/)
- [Freie Schule Salzburg](https://freieschulesalzburg.at)
- [When is criticism a gift and when is it just fear?](https://zukunftbilden.eu/en/denken/kritik-als-geschenk)
- [AI implementation for companies](https://zukunftbilden.eu/en/ki-umsetzung)

## Margin notes

- **Randnotiz** The rule to cut applies to every tool, even proven ones. Effort, habit and pride do not count as a reason to continue.
- **Verknüpft**  [The first agent is always the guardian of the brand.](https://zukunftbilden.eu/en/denken/waechter-der-marke), [When is criticism a gift and when is it just fear?](https://zukunftbilden.eu/en/denken/kritik-als-geschenk), [Paideia for schools](https://zukunftbilden.eu/en/paideia), [AI implementation for companies](https://zukunftbilden.eu/en/ki-umsetzung)
- **Beleg** Paideia core handbook, chapter 20 (The school that is never finished), section 20.5 of 13 August 2026, sections 20.1, 20.3, 20.4 and 20.6, status September 2026. Cymbal decision paper of 13 August 2026. Origin of kaizen and first principles following the school’s own article, see sources. Alongside it a private video from 2018 ("Scripter for Beta Testers").

## How this note grew

- 2026-09-03: Deepened with an example of my own from 2018: a tool I had built automated before it was robust, exactly the mistake this order prevents today.
- 2026-09-03: Grown: the five steps in the original wording, the sourced origin in kaizen and first principles, the cymbal from hand to striking robot, the guiding questions with the delete list as a tool. The earlier attribution to Elon Musk is gone, because the handbook does not contain it.
- 2026-09-02: Planted from the improvement algorithm in the handbook.

---

Source: https://zukunftbilden.eu/en/denken/loeschen-vor-automatisieren · Feed: https://zukunftbilden.eu/en/denken/feed.xml