10. THE VOICE REGAINED

Building a proprietary dataset is the foundation of our reappropriation.

I mentioned earlier the use of such data in commercial AI models, but I urge the greatest caution. Like other products from Big Tech, commercial AIs are only too eager to exploit your data. Yet your dataset is your gold mine. It must be jealously guarded. To reduce the risks, one can, for instance, access models through an API (on OpenRouter, for example); the user is then anonymized and, depending on the policies, the data may be protected. In the long run, I would like to be able to use an open-source AI running on my own computer to work with my data corpus, but for now, 1) commercial models remain superior, and 2) running the best open models locally — Kimi K2.5, for example — requires a very powerful computer. For literary creation, where AI still struggles, you absolutely must use the cream of the crop; otherwise you will only be stitching together clichés.

I am in the process of building my own tool: this is possible today using Claude Code or other AI coding agents. The architecture will include a RAG — Retrieval-Augmented Generation — that is, a connection between the brain (the AI) and a personal library (my complete or curated data corpus). This technique goes far beyond the enriched prompt I demonstrated earlier, because it allows the AI to draw on the proprietary data mine as it prepares its response.

If, as I believe, the dataset will enable us to find the voice or voices — the vocal imprint — then this step will mark something like our Time Regained. I am sure that some readers have been thinking: “Find one’s voice? How dreadful! What matters is the journey. If you lay hands on the Golden Fleece, the quest is over.” That is true — and this is precisely what is unsettling about AI, or at least about the prospects that artificial intelligence opens up. I touched on this when I discussed Claude Code: there is something almost magical in AI’s capacity — and above all, its future capacity — to project us directly into what we normally call the “end” (fulfillment, resolution).

The end, as we know, is only the beginning of something else. Time Regained does not exactly mark the conclusion of the Search but its imagined recommencement — the start of a new cycle, of a rewriting or a rereading — shifted, however, because we have already arrived. The same will hold for voice. Once we have found our imprint, one might say it is all over. There is no longer any need to write. In fact, what I foresee is the beginning of a new cycle. What will remain is to actualize that voice, to give it a thousand bodies, a thousand texts, a thousand lives in the new world. On the far side of the end, a renewed story opens up, one that deviates from the narrative models we know. It will no longer be the quest; it will be the request. An actualization of forms after the resolution. The recommencement of the quest, but in the universe of information. The request — that is, the prompt that relaunches the quest once the voice has been elucidated.

“But voice is not mathematical,” some will object. “You cannot find the voice by means of data analysis, however intelligent.” Perhaps. More likely, we will find precise and granular imprints that, while not definitive, will serve as matrices from which a multiplicity of forms can be generated. I know full well that we write with our bodies; I know there is something organic in the grain of the voice. But what if even the body (and its motions: fear, desire, flight, and so on) could, in the final analysis, be reduced to a skein of connections — an entanglement that does not sum up voices but carries their potentialities? By descending deep enough into the granularity of voice, we may well find mathematics.

This possibility wounds our human narcissism: we would like to believe ourselves irreducible, but are we really? The notion of irreducibility appears rather like the last bastion of a transcendence we have never managed to purge from our thinking. We may be complex, yes, but if we are not irreducible, we find ourselves cast onto the same plane as everything else: that of data. Yes, this includes the body; yes, this includes emotions and intuition. In the age of intelligence, we are a body of data.