From the application form

Aufhebung: keep, move past, lift

Our application form asks for Aufhebung. This page explains what the word means, how we use it at Gradiva and what it looks like in real industries. You do not need any philosophy to apply; the three moves below are enough.

If you came here from the application form, it is still open in the other tab.

Careers page To the application form Türkçe

One German word, three moves

Aufhebung is the noun of the German verb aufheben. In everyday German the verb has three meanings that seem to pull against each other:

Keep
einen Brief aufheben: to keep a letter, put it away safely
Keep what works, on both sides. In the result you should be able to point to where each side’s strength now lives.
Move past
ein Gesetz aufheben: to repeal a law, end it
End what does not work. Something has to go; a result that keeps everything is a pile, not a step.
Lift
einen Stift aufheben: to pick a pen up off the floor
Bring the two together at a higher level. The result can do something that neither side could do on its own.

The philosopher G. W. F. Hegel turned this everyday word into a working term. In his Science of Logic (first published 1812) he notes that aufheben means both to preserve and to put an end to, and that it is a pleasure for thought to find one word holding opposite meanings. In his use, when two positions clash, the way forward is neither to pick one nor to split the difference. It is a third position that ends what was wrong in both, keeps what was right in both, and stands one level higher than either.

One level higher

kept A’s strength
kept B’s strength
new something neither had

Side A

kept what works
ended what doesn’t

Side B

kept what works
ended what doesn’t

What an Aufhebung keeps, ends and adds.

In English the term is usually translated as sublation. We keep the German word because no single English word holds all three moves.

Three things that look like it but are not

Picking a side. Choosing A over B ends B but keeps nothing from it. You have won an argument, not moved past it.
Splitting the difference. Half of A and half of B is a compromise. Each side loses something and nothing new appears. A price halfway between cheap and premium often pleases neither buyer.
Renaming. Calling the clash “a balance” or “both/and” leaves every problem where it was.

A quick test we use on our own drafts: take the word Aufhebung out of your sentence and say plainly what changes. If what is left is “balance both”, “it depends” or “both are valid”, the work is not done yet.

Four questions we ask at Gradiva

We use Aufhebung as a working tool: in product decisions, in design reviews and when we judge our own work. Before we call something an Aufhebung, it has to pass four questions.

Has something really ended? Could either side still say, “I was enough on my own”? If so, nothing was moved past; it was only renamed.
Where did the good parts go? Point to the exact place in the result where each side’s strength lives. If you cannot, something was lost and called progress.
What can it do that neither side could? Name one new thing. If the result is just A plus B, it is bigger, not higher.
Does it run one way only? Try to go back: rebuild the two old positions from the new one without losing anything. If that is easy, you have described the same thing twice instead of moving forward.

One more habit: we write a deliberately vague answer that sounds similar (“combine the best of both”) and put it through the same four questions. If the vague answer passes too, we asked the questions too loosely.

Not every clash needs one

Sometimes the right move is to see that the clash is not real. We check three cases before reaching for Aufhebung:

Same thing, two names. “The glass is half full” and “the glass is half empty” describe one glass. There is nothing to lift; pick the description that helps your reader.
Two different questions. “Should we write it in Python, or should it be fast?” mixes two questions: which language, and how fast it needs to be. Separate them and answer each one.
Hidden options. “Raise the price or lower it?” hides other moves: change what the price includes, offer tiers, change when people pay. Open up the options before you choose.

Aufhebung is for the case that remains: two positions that are each partly right, where neither can simply win.

Examples from industry

Except for the first, none of these teams set out to “do Aufhebung”. We read their solutions through the four questions because they show the shape clearly. Each example follows the order of those questions: what ended, what was kept, what lifted both.

Film

Montage: a third meaning from two shots

Where does a film’s meaning live: in each shot the camera records, or in the order the editor builds?

Move past: the idea that a shot carries its meaning alone.
Keep: both shots, untouched; nothing inside them changes.
Lift: side by side they produce a third meaning, such as hunger, that neither shot contains.

Lev Kuleshov showed this around 1920: the same blank face of an actor, cut next to a bowl of soup, a coffin or a woman on a sofa, was read as hunger, grief or desire. Sergei Eisenstein then built his theory of montage explicitly on this line of thought: for him a cut is a collision of two shots, and from the collision a new concept arises. Limit: Kuleshov’s original footage is lost and modern repeats of the experiment give mixed results; editors rely on the editing principle, not on the exact experiment.

Pricing

Versions: let the buyer choose the price

A low price brings many buyers; a high price brings margin. One price cannot do both.

Move past: one price for everyone.
Keep: the low entry price and the high-margin offer, side by side.
Lift: tiers let each buyer pick their own level, so the price list itself sorts customers by what they value. No single price can do that.

Carl Shapiro and Hal Varian described this as versioning in 1998; it is why so much software sells as Basic, Pro and Business. Limit: badly designed tiers turn back into a compromise, with every tier a little too expensive for the people it is meant for.

Manufacturing

Toyota: stop the line so it stops less

Keep the line moving to hit output, or stop it to fix defects?

Move past: the belief that every stop is a loss.
Keep: high output, and no defects passed down the line.
Lift: any worker can stop the line at a problem; the cause is fixed where it starts, so over time the line stops less and quality rises with output.

This is the jidoka principle and the andon cord of the Toyota Production System, described by Taiichi Ohno. Limit: it works only when a stop is followed by real root-cause fixing; without that, a stop cord is just a slower line.

Product

iPhone: a keyboard that appears when needed

In 2006 phones chose between a physical keyboard for typing and a large screen for viewing.

Move past: fixed plastic keys that sit there whether you need them or not.
Keep: typing, and a full screen.
Lift: a keyboard drawn in software: it shows up only when you type and changes with the app and the language.

Steve Jobs set out the clash himself when he introduced the iPhone in January 2007. Limit: something was lost too, the feel of physical keys, and some users missed it for years. A real Aufhebung can still have a cost; the question is whether the gain is greater.

Software

Speed and stability together

Ship fast, or keep systems stable? For years teams treated this as a trade-off.

Move past: big, rare, risky releases.
Keep: speed and stability, both.
Lift: small, frequent, automated releases. Each change is small enough to check and undo, so going faster also makes the system safer.

Several years of survey research by the DORA team, published as Accelerate (2018), found that the strongest teams lead on speed and stability at the same time. Limit: the research is survey-based and shows an association across many companies; it does not prove that any one team will gain by copying the practices.

A small worked example

A bakery has two posters for its window. The first lists every loaf with ingredients and prices: accurate, but nobody stops to read it. The second is a large close-up of bread: people stop, but they do not know what to buy.

Picking a side: use the photo. The information is lost.
Splitting the difference: half photo, half list. Both get smaller and neither works.
Aufhebung: one close-up of today’s loaf, with only its price and the time it came out of the oven.

Run it through the four questions. Ended: the long list and the anonymous picture. Kept: the photo’s pull and the list’s facts. New: freshness becomes the message, something neither poster said. One way only: from the new poster you cannot go back to the two old ones without losing that message.