A Candide Without Voltaire?

With artificial intelligence, a work can become far more elaborate than its creator’s own resources might lead us to expect. That possibility forces us to reconsider what works tell us about their authors—and how we distribute merit.

The controversy surrounding Thélyson Orélien’s C’était ça ou mourir raises a question that reaches beyond this novel: what does a work establish about the person who signs it? This question extends an earlier reflection here on access to writing with AI.¹

The case itself calls for precision. On October 21, 2026, Éditions du Boréal announced that it was ending its collaboration with the writer, saying it had found traces of AI use in the novel’s working documents. The publisher nevertheless stated that it did not know the extent of that use.² Orélien, for his part, maintains that he wrote the book and says he used these tools for research.³ These positions do not allow us to reconstruct the precise contribution of each intervention here.

The use of a tool, the honesty of statements about that use, and the value of a text call for distinct assessments. We can ask for transparency without declaring that a work created with AI assistance must be uninteresting. We can also acknowledge that a text moves its readers without treating that response as an answer to every question about how it was made.

Let us therefore leave the evaluation of this novel open and consider a broader hypothesis. Could someone, with substantial assistance, produce a significant work without personally possessing all the knowledge, mastery or intellectual depth that we might be tempted to attribute to them on reading it?

The title of this essay takes that hypothesis to its limit: a Candide without Voltaire, referring to the philosophical tale published in 1759.⁴ It does not assume that current models could write an equivalent of that tale. It asks how the importance of a work relates to the stature of the person who claims its creation.

What We Think We Recognize in a Work

Faced with a remarkable achievement, we readily reason backwards from the result to the person. A subtle book suggests a subtle mind. A complex program suggests substantial expertise. An accomplished painting bears witness to years of practice.

This reasoning has never been infallible. Books have editors, studios have assistants, software has libraries and teams. The name in the foreground does not tell the whole story of production. Individual abilities have always depended on collective resources.⁵

AI can, however, widen the gap between what the result displays and what the person signing it actually did. It allows us to delegate operations that we associated with competence itself: developing an argument, proposing a structure, formulating an objection, writing a scene or producing code.

A sophisticated work remains evidence of abilities brought to bear somewhere. It tells us less directly how those abilities were distributed. Its quality alone no longer establishes that the person presenting it could explain how it works, recognize its limitations or repeat the performance.

This uncertainty touches a deep social habit: we award positions, authority and reputations on the basis of what people produce. If the relationship between results and individual abilities changes, those judgements need to be reconsidered.

Conception, Execution, Judgement

To clarify the issue, let us distinguish three activities. Conception means setting a direction, framing a problem, imagining a form. Execution means bringing that form into being in a medium: language, colour, sound, code. Judgement means recognizing what works, discerning what is missing and deciding what deserves to be kept.

These activities constantly overlap. Writers discover what they think by writing. Painters change their intentions when a colour resists them. Programmers understand a problem better by trying to solve it. Execution produces ideas too; it does more than put them into finished form.

AI assistance can redistribute these activities. Someone can define an intention, request several proposals, reject some, transform others and direct a long sequence of revisions. A real contribution then lies in directing the work, even when many operations are delegated.

But it would be too convenient to conclude that judgement can effortlessly replace every skill that has become less necessary to carrying out the work. Where does the ability to recognize a weakness come from? Often from having encountered the problem ourselves. We can appreciate a passage without understanding its implications, or select a software solution without seeing its fragility.

Judgement can itself be assisted. We can ask a system to assess its proposals, compare them and correct them. That can improve the work. It can also leave users dependent on an assessment they do not know how to challenge.

The decisive question then becomes: which decisions did the person actually make, with what understanding, and with what ability to disagree?

From the Paintbrush to the Generated Image

The analogy with visual art helps clarify the shift. Painting on canvas, working on a digital image and directing AI generation can produce equally striking images. The activities and skills involved are nevertheless different.

Traditional painting involves gesture and knowledge of materials. Digital art can require a deep command of drawing, composition, light or image manipulation. Among other things, it adds the ability to undo an operation, layer elements and automate certain processes. Moving to digital tools therefore does not, in itself, make craftsmanship disappear.

Generative AI allows more of an image’s construction to be delegated. A description can receive a visual interpretation whose details the user did not determine. But practices range from immediately accepting a proposal to demanding work involving research, selection, compositing and retouching.

The tool’s name alone tells us little about this diversity. To attribute merit, we need to examine decisions and work. To appreciate an image, we also need to look at the image. And to recognize a particular virtuosity—in drawing or painterly gesture, for example—we need to know whether it was actually exercised.

A strong image can thus reveal an eye without demonstrating the mastery of the paintbrush that its surface seems to suggest. It can also depict a world largely proposed by the system, in which the user mainly recognized something that resonated with them.

The Double Paradox

An ordinary idea can now receive an impressive form. A conventional theme can become a fluent narrative, a vague proposal can display all the signs of rigour, and a rudimentary intuition can take on the appearance of a finished project.

The result may even be a real success, containing discoveries that the initial request gave no reason to anticipate. Users can then receive as much as they contribute: they recognize a success without having conceived all that makes it work.

The risk lies in confusing these qualities of presentation with what they appear to guarantee. Visible sophistication can lend credibility to an idea that has not been developed in depth. It can also give a person credit for abilities they did not have to exercise.

But the reverse movement deserves just as much attention. An original idea can remain trapped for a long time by insufficient technical skill. Someone can know a situation intimately, imagine a useful application or carry a distinctive experience without having the means to give it a shareable form.

Assistance can then open a way through. Consider a community worker who imagines a tool suited to the people she supports, without knowing how to program. Or someone who wants to convey an experience in a language they have not fully mastered. That technical gap negates neither their knowledge of the situation nor the originality of what they are trying to express.

It would nevertheless be an illusion to believe that an idea always arrives complete, awaiting execution. Trials, failures and the resistance encountered in making something help shape it. Assistance can make those explorations more accessible; it can also bypass them if we accept the first convincing form too quickly.

The same tools can therefore amplify the conventional and make something new possible. The challenge is to understand the conditions under which each occurs.

Software That Works—What Comes Next?

Software makes this distinction particularly tangible, extending the reflection developed here on the decisions a prototype can inform.⁶ Imagine a small association that, with AI assistance, obtains a registration application meeting a need previously beyond its reach. The benefit is real: a problem has been understood and a solution becomes usable.

The application’s existence does not, however, demonstrate that its initiator could protect its data, diagnose a failure or develop it further. A successful demonstration provides information about the functioning observed. It does not prove every ability required for sustained use.

We need to recognize both the merit of identifying the need and the skills still required to carry the work forward. Demanding that every initiator be able to rebuild everything alone would be absurd: professionals themselves work with dependencies and collaborators. Trust also rests on the ability to understand enough, verify, seek help and take responsibility for what is delivered.

This reasoning applies to artistic works, with different criteria. A work’s significance is a separate question from its author’s personal virtuosity. It also applies to organizations: an impeccable report can circulate without anyone being able to defend its assumptions.

What Institutions Will Need to Recognize

Schools, hiring processes, prizes and organizations all assign value to achievements. Yet they are not all seeking to recognize the same thing.

A school assignment can serve to develop and assess a personal ability. A professional tool may be judged first on its usefulness and reliability. An artistic prize may seek to distinguish a work, an approach or a contribution to a tradition. Rules governing delegation should take these purposes into account.

When the aim is to establish what someone understands, a finished result may sometimes need to be accompanied by a conversation, a revision or an explanation of choices. When the aim is usefulness, we will need to examine the work’s effects and the ability to take responsibility for it. When the aim is artistic, transparency about the process can inform attribution without replacing the encounter with the work.

We will also need to prevent requests for evidence from becoming permanent suspicion, applied more harshly to some people than to others. We can make a contribution intelligible without requiring surveillance of every creative act.

Finally, access to tools is not enough to ensure equality. Available time, financial resources, support and distribution networks continue to matter. And the capabilities made accessible themselves depend on infrastructure and human productions whose recognition remains contested. The contribution of the person signing the work needs to be situated within that larger whole.

A Work, a Contribution, an Authority

We can distinguish three questions: does this work have value? What did the person signing it contribute? Which abilities does it give us grounds to recognize in them?

The answers may diverge. A work can be important without having been executed from beginning to end by the person whose name it bears. A contribution can be original despite limited technique. And a brilliant production can warrant interest without granting its author general authority.

We will need to preserve places where people learn by doing, where the resistance of materials shapes thought. We will also need to welcome creations made possible by new means. These two commitments can coexist: access to expression and the development of abilities do not meet exactly the same need.

Voltaire then brings us back to the scope of the initial question. If someone produced a remarkable tale with substantial assistance, we could recognize the text’s strength and seek to understand their contribution. It would remain unwarranted to attribute to them, on that basis alone, the full breadth of thought we associate with Voltaire.

A work can be worth more than what it proves about its author. Recognizing that gap might help us welcome new creations—and distribute our admiration, our trust and our power more fairly.

Editorial note: this essay grew out of Stéphanie Savard’s reflections and exchanges with ChatGPT. The starting point, the thesis about the relationship between an achievement and its author’s abilities, the visual arts analogy and the double paradox came from the author. The AI assistant contributed to the structure, writing, development of the reasoning, revisions, source research and English translation. Its intervention therefore extends beyond stylistic presentation alone. The sharing cards were generated with ImageGen. The community worker and association examples are hypothetical.

Sources and Further Reading

The factual sources below document the positions expressed in the controversy and the reference to Candide. Links to Bifurcations essays situate the reflection; they are not independent evidence for the thesis. This essay offers neither an expert examination of the novel nor an assessment of its author’s abilities.