Peer review is often presented as the threshold through which knowledge acquires legitimacy, yet its authority rests on a far more contingent architecture than the institutional language surrounding it usually admits. A manuscript is commonly judged by a small number of individuals selected by an editor according to disciplinary proximity, publication history, availability, reputation or recommendation; these reviewers may possess exceptional command of the subject, partial familiarity with it, or expertise confined to a neighbouring paradigm, and their assessments inevitably carry the limits of their formation, workload, institutional position and intellectual commitments. The system therefore does not embody an impersonal intelligence capable of recognising truth, but a provisional arrangement of delegated judgement. Its value lies in organised scrutiny: an argument is exposed to criticism, evidence is tested, omissions are identified and claims may be strengthened before publication. Its weakness begins when this useful procedure is confused with an absolute measure of originality, validity or historical importance. Two anonymous reports cannot exhaust the significance of an unfamiliar work, particularly when that work crosses disciplines, constructs its own vocabulary or challenges the assumptions by which the journal itself distributes relevance. Reviewers are rarely neutral observers standing outside the field they assess; they are participants in its hierarchies, schools, conventions and economies of attention. They may reward methodological recognisability, textual obedience and citational belonging while treating conceptual displacement as confusion, institutional independence as insufficient legitimacy or transdisciplinary ambition as a failure to remain within jurisdiction. The result is not simply bias in the psychological sense, but a structural conservatism produced by the geometry of selection: editors choose recognised specialists, specialists apply inherited criteria, and the publication system converts those criteria into visible authority. Peer review can therefore improve a text while simultaneously narrowing the conditions under which new forms of knowledge become admissible. Its authority is further complicated by scarcity. Review is unpaid or weakly recognised labour performed under pressure, often by researchers already burdened with teaching, administration, funding applications and publication demands. The ritual may appear rigorous at the institutional scale while remaining hurried at the level of actual reading. Some manuscripts receive attentive engagement; others are filtered through fatigue, haste or disciplinary reflex. The phrase “peer reviewed” conceals this variability by transforming an uneven encounter into a stable credential. What the credential certifies is not truth but passage through a recognised procedure. This distinction matters because contemporary knowledge no longer develops exclusively through discrete articles entering closed journals at measured intervals. Large open corpora, persistent repositories, version histories, linked bibliographies and computational methods make possible another model of evaluation: continuous, cumulative and revisable rather than occasional, concentrated and terminal. In such a model, criticism is distributed across time and across different forms of intelligence. Human readers remain responsible for interpretation, conceptual judgement and historical understanding, while computational systems can assist by detecting recurrence, inconsistency, semantic drift, missing relations, bibliographic imbalance and structural redundancy across quantities of material inaccessible to unaided reading. Large language models are significant here not as surrogate authorities but as instruments of distinction. Their value emerges when they are used repeatedly, comparatively and critically to expose differences that might otherwise remain submerged within scale. They can sharpen formulation, identify weak transitions, compare variants, trace the recurrence of concepts and reveal where apparent diversity conceals repetition. Their judgements remain conditioned by training data, prompts and editorial framing, yet the same is true, in different form, of every human reviewer. The essential question is therefore not whether evaluation is human or computational, traditional or alternative, but whether its procedures are visible, contestable, plural and capable of revision. A corpus containing thousands of nodes, bibliographic relations, external agents, persistent identifiers and documented transformations can establish a form of epistemic accountability that differs from journal certification. Its legitimacy does not arise from a single moment of institutional acceptance but from the cumulative density of its structure: claims can be located, compared, corrected, contextualised and challenged; concepts can be followed through successive versions; exclusions become detectable; and editorial decisions leave traces rather than disappearing behind anonymous verdicts. Scale alone would remain inert, but scale organised through naming, indexing, recurrence and verification becomes a material force. It creates a field in which authority is not bestowed once but continually tested. This does not abolish the need for expert criticism, nor does it guarantee that an open system is free from error, vanity or self-confirmation. It changes the location of judgement. Instead of allowing two or three selected readers to function as a final gate, evaluation becomes an extended public architecture composed of many encounters, revisions and forms of evidence. Such an architecture may prove especially important for work produced outside universities, journals and conventional research careers, because it demonstrates that intellectual seriousness can be constructed through persistent labour, methodological clarity and exposed infrastructure rather than inherited affiliation alone. The deepest challenge posed to traditional peer review is therefore not that its reviewers sometimes know too little, but that the system grants disproportionate symbolic power to a narrow and opaque episode within the life of an idea. Knowledge deserves more than admission or rejection at a single door. It requires environments in which propositions can remain available, revisable and answerable to readers not yet selected, disciplines not yet stabilised and questions not yet recognised by the institutions appointed to judge them.
