The Difference Between What Is Said and What Has Become Thinkable
It happens, at the end of an intense conversation, that one says: “we touched on something important without formulating it.” The idea was there, carried by the exchanges, prepared by the successive contradictions, but no one said it. It remained on the horizon — visible, untraversed.
This experience is not anecdotal. It points to a precise phenomenon: a debate produces more than it says. It makes certain formulations possible that were not so before it, even if no one actually proposed them. It opens up a conceptual space of which it explores only a part. It opens questions it does not pose.
Horizon of Possibilities is the analysis mode designed to identify precisely this layer. Not what was said (that is the work of Integrative Synthesis), not what was built through the exchange (that is the work of Emergence Analysis), but what the debate made formulable without formulating it.
A Deleuzian Echo
The gesture of Horizon of Possibilities finds a conceptual echo in Gilles Deleuze, and the reference deserves to be evoked briefly to situate what the mode does on a more theoretical plane.
Deleuze distinguished two notions one tends to conflate: the possible and the virtual. The possible is what could come about by following the existing order — a different combination of already-given elements, another distribution of what is. The virtual is of an entirely different nature: it is what insists in the real without being actualized in it, what haunts it as an undeployed disposition. A debate that circles around an idea without ever formulating it makes that idea virtually present — it operates on the debate (orienting its tensions, preparing its contradictions) without appearing in it as a statement. Horizon of Possibilities seeks to identify and formulate these virtualities: not combinatorial possibles one could have constructed, but insistences that were already there, carried by what was said but not actualised.
This reference, like the Foucauldian reference in the Critical Archaeology mode, is not a claim of direct lineage. The Horizon of Possibilities mode is not an application of Deleuzian philosophy. But the actual/virtual distinction provides a precise vocabulary for describing what the mode does, and this precision can help one understand why Horizon of Possibilities does not produce free speculation — it operates under a constraint of immanence: every formulated virtuality must be anchorable in the debate that actually took place.
An Operation of Formulation in the Debate’s Place
Concretely, the mode performs a particular gesture. It re-reads the session produced by Metamorfon and enunciates in the models’ place what they had prepared without saying. Not what they could have said had they been given other instructions — that would be arbitrary speculation. But what was already present in the negative in their exchanges, in the form of openings, unresolved tensions, undrawn implications.
For this operation to be rigorous, it must respect a strong constraint: every horizon identified must correspond to a real passage of the debate. It is not the mode that invents leads; it is the mode that completes leads the debate had opened without traversing. The difference is decisive. It distinguishes Horizon of Possibilities from creative production with no supporting material.
Three types of discoveries recur regularly in what the mode produces.
First Type: Concepts a Word Away
It happens that a debate approaches a concept without naming it — that it circles around it, prepares it through its contradictions, but stumbles for lack of a vocabulary. Horizon of Possibilities identifies these concepts and formulates them.
Example from a session on AIs that speak like humans: the two models opposed the human’s bodily vulnerability to the traceability of machines. Horizon of Possibilities noted that the debate was approaching a third path without naming it — the criterion of the deliberating citizen could be the capacity to be sanctioned over time. Neither embodiment nor traceability: a temporal property of accountability. This formulation was in no turn, but it became formulable after reading the turns.
Second Type: Structural Displacements of the Problem
It happens that a debate, in handling its question, tacitly displaces the question itself without noticing. The interlocutors begin by interrogating one object, and over the course of the turns, another object progressively takes its place without anyone thematizing the substitution. Horizon of Possibilities names this drift.
In a session devoted to the question “What does a democratic society lose when machines speak like humans without signalling their artificial nature?”, the debate began by inquiring into the loss of epistemic integrity, the dissolution of a communicational contract, the neutralization of minority voices. But progressively, over the course of critical and refutational turns, the object of the debate shifted: the discussion was no longer really about what a society loses, but about what obligations to impose on technical infrastructures. Horizon of Possibilities brought to light that this drift signalled something important that had not been said: democratic sovereignty is shifting towards material layers (GPUs, data centres, networks), which redefines the constitutional question itself. This formulation was in no turn; it nonetheless captured what the dynamic of the debate had produced without naming. Once formulated, it reoriented the reading of the entire preceding session.
Identifying this type of displacement has practical value. It allows one to know which question one is actually treating, which is not always the question one thought one was treating. For a user preparing a decision on the basis of a debate, this is decisive information.
Third Type: Questions the Debate Prepares Without Posing
Finally, certain questions become evident after a debate, when they had been invisible before it. The exchanges, in exploring a conceptual space, render formulable enquiries that would have seemed premature or ill-posed at the outset. Horizon of Possibilities identifies them and poses them.
In the same session on machines that speak like humans, Horizon of Possibilities posed in closing a question that could not have been posed at the start: if the normalization of synthetic speech erodes the very category of the “authentic” faster than any regulation — such that no signaling will soon be able to produce any real epistemic effect — then must democracy protect the human/machine distinction, or learn to deliberate without it, founding its legitimacy on something other than the provenance of speech? This question is striking for three reasons. It was not formulable at turn 0: it presupposes the exhaustion of the regulatory responses the debate had just explored. Nor is it a summary of what was said: the models had neither admitted this erosion nor suggested learning to deliberate otherwise. It works from what was said to open a horizon neither interlocutor had reached.
It is precisely this type of question that makes Horizon of Possibilities an instrument of prospective thought, and not a mere summary dressed up as speculation. The debate, exhausting its own resources, prepared a question it could not pose itself. Horizon of Possibilities formulates it, and that formulation opens a space of reflection that did not exist before the debate took place.
What Distinguishes Horizon of Possibilities From Other Modes
A risk of confusion must be cleared up, because it bears on the most subtle distinction among the seven modes. Horizon of Possibilities is not Emergence Analysis.
Both modes deal with what is new in a debate. But they operate on different layers of novelty.
Emergence Analysis maps the concepts that actually appeared in the exchanges — ideas that were in no starting point (neither initial prompt nor user intervention) but that were produced and formulated by the dialogue itself. These concepts are statements attestable in the corpus: they can be cited, located in a precise turn, and identified as introduced by a particular model. “Legitimacy debt” in a session on citizen consultations, “revocable infrastructure” in a session on machine speech, “axiological fork” in a session on AI ethics: these concepts did not pre-exist the debate, but they figure explicitly within it once it has taken place. Emergence Analysis records them and traces their path.
Horizon of Possibilities, by contrast, maps what remained on the horizon: what was prepared by the debate but not enunciated. The concepts it formulates are not in the corpus; they must be sought there as virtualities, not as actualities. They were not produced by the interlocutors; they were carried by the argumentative dynamic without being grasped. When Horizon of Possibilities names “democratic sovereignty is shifting towards material layers” or “learning to deliberate without the human/machine distinction”, these formulations appear in no turn. They formulate what the conjunction of turns made thinkable.
The two modes complement each other because they operate on two different layers of novelty: the actual (Emergence) and the virtual (Horizon). To use one without the other is to risk missing half of what the debate produced. Emergence Analysis identifies the new concepts that were gained; Horizon of Possibilities identifies the new questions there was no time to pose.
Another distinction deserves brief mention. Horizon of Possibilities is also not Tension Mapping. Tension Mapping charts what resists in the debate — the irreducible divergences, the disagreements that even refutational mode did not resolve. Horizon of Possibilities is not interested in what resists but in what opens. In some cases the two modes can converge (an unresolved tension can be the seed of an unformulated virtuality), but the analytical gesture remains different: one fixes what does not move, the other anticipates what could move.
When to Use Horizon of Possibilities, When to Set It Aside
Horizon of Possibilities is probably the most demanding mode for the model that conducts it, and the most delicate to read. It is not the first mode to activate on a session, nor one that can be deployed indifferently.
A few situations where it is particularly relevant.
When a session is complete and one needs to draw directions for extension from it. A Metamorfon session rarely exhausts its own object. Horizon of Possibilities identifies what the session prepared without exploring, and thus naturally indicates where a continuation might unfold — a new session on a reformulated question, an article extending the debate, a personal reflection to pursue. For anyone using Metamorfon as a research or creative tool, this is a generative mode.
When seeking new questions on a saturated subject. Some subjects appear to have been treated from every angle — AI ethics, deliberative democracy, algorithmic governance. Horizon of Possibilities is valuable in such cases because it formulates what even the richest debates have left at their periphery. The questions it produces often have a conceptual freshness that the debate itself had not reached.
When one wants to avoid stopping too early at an apparent consensus. When a debate seems to arrive at convergent conclusions, Horizon of Possibilities can bring to light that this convergence leaves open problems the interlocutors did not know how — or did not want — to name. It is a useful counter-poison against premature closures.
When preparing editorial or intellectual content. For a researcher, journalist, essayist, or consultant writing on the basis of a debate, Horizon of Possibilities points to the original angles — those that were not treated, those still to be formulated. It is a particularly fertile instrument upstream of a written production.
Conversely, a few situations where Horizon of Possibilities is unsuited.
When one simply wants to understand what was said. An Integrative Synthesis will largely suffice. Horizon of Possibilities supposes that one already has a good understanding of the content; read first, it can be disorienting.
When the debate has been brief or weakly polarized. Horizon of Possibilities needs rich material to operate. On three or four turns in balanced mode, it has little virtual matter to formulate, and risks producing vague or generic observations.
When one is seeking a factual, verifiable, traceable analysis. By its very nature, Horizon of Possibilities formulates virtualities — hence formulations that are not in the debate. They must be anchorable in what was said, but they are not quotations. For a use where literal traceability is critical (legal analysis, strict methodological audit), other modes will be more appropriate.
It should finally be noted that a good Horizon of Possibilities requires a powerful model to conduct it. Smaller models risk producing insufficiently anchored speculation — combinatorial possibles they present as virtualities. The Deleuzian distinction evoked at the outset is not a theoretical comfort: it has precise operational consequences. For this mode more than for the others, the choice of analysis model matters.
What Horizon of Possibilities Brings Beyond Its Results
There remains, in closing, something the previous sections may have left implicit. Horizon of Possibilities does not produce only useful observations at the end of a session. It also produces a shift of gaze on reading itself. Once one has read a Horizon of Possibilities on a session one knows, one re-reads the debate differently: one becomes attentive to what the interlocutors touch on without naming, to what their contradictions prepare without formulating, to what their convergence leaves open. This gaze is not reproducible on demand on just any debate, but it is acquired progressively through the practice of sessions analysed under this mode.
In this sense, Horizon of Possibilities is perhaps less an output to be consumed than an exercise that forms an intellectual disposition. It is a property it probably shares with the other Metamorfon modes, but one that is particularly salient for this mode, because the object it treats — the virtual, the insistent, the unformulated — is precisely what one learns to see by practicing it.