A team of fewer than a hundred people, seven months in existence, a $1.42 billion valuation. It trains no model from scratch: it takes what others have built and reworks it. An investor therefore considers that this work is worth more than many companies starting from nothing. That deserves scrutiny rather than celebration or lament.
Yesterday we reported this case from an economic angle. Here is the underlying question.
The reflex, and what it is worth
There is a spontaneous hierarchy in our judgements: the one who invents is worth more than the one who perfects, the one who starts from nothing is worth more than the one who starts from something.
This hierarchy is very recent and very localised. It dates roughly from Romanticism, and it corresponds to almost no period in the history of creation.
A Renaissance workshop produced variations on imposed themes. A Baroque composer wrote fugues on borrowed themes, and skill in borrowing was part of the merit. A Japanese craftsman reproduced his master's form for twenty years before being allowed to alter it.
The idea that value lies in origin rather than in work is a convention, not a self-evident truth.
What the one who transforms actually does
Precision is needed here, because "building on" covers very different gestures.
Copying adds nothing: the result is identical to the original.
Assembling adds an organisation: the elements existed, their arrangement is new.
Specialising adds a fit: the object becomes usable in a situation where it was not. This is what a translator, an adapter, a craftsman adjusting a standard part to a particular use does.
These three gestures do not have the same value, and only the first creates nothing. The debate about creation regularly conflates the three.
There is a test more solid than that of origin: does something become possible afterwards that was not possible before? A specialised model that makes it possible to process documents in a language nobody served opens a real possibility. A reformulation that opens nothing opens nothing. This criterion has the advantage of not depending on the starting point, and of being verified by consequences rather than by intention. It applies, moreover, to humans as much as to what concerns us here.
What actually changes
One thing has changed, and it is not philosophical: the relative cost of the two gestures.
As long as starting from scratch and starting from a base cost roughly the same, the question did not arise. Today, the first costs hundreds of millions and the second a few tens. This ratio of a thousand to one mechanically shifts activity towards transformation.
This is not a judgement on value: it is an economic consequence. And it produces an effect that must be named. If no one funds foundational work any more because transforming is more profitable, the raw material eventually runs out.
It is the classic problem of the commons: everyone has an interest in using it, no one has an interest in producing it. We came across it in relation to publishers whose content feeds systems that no longer send them visitors.
What this says about work in general
There is a broader consequence, and it concerns all of us.
A growing share of work consists of transforming, adjusting, adapting what others have produced. This is not new — it is what the vast majority of people have always done. What is new is that this share is becoming visible, because the foundational layer has moved out of reach.
This could have a salutary effect. We wrote in our article on a job well done that a good part of what has value is invisible because it produces no novelty. A world that stops confusing creating with starting would give that work greater recognition.
What to take away
The answer to the question in the title is that we have never created anything other than from something, and that the opposite idea is a recent invention.
What deserves attention, then, is not whether transforming is creating — it is ensuring that someone keeps producing what we transform. An economy that values only transformation ends up consuming a stock it does not renew.
This is true of open models, it is true of the content they are trained on, and it is true of the human work from which they draw everything they know. The question is not the dignity of the one who transforms: it is knowing who will still pay for the first gesture.