Matthew Kelly’s Situated Epistemic Infrastructures proposes a diagnostic framework for understanding knowledge after the collapse of stable epistemic coherence. The paper argues that large language models expose, rather than merely cause, a deeper crisis in contemporary knowledge systems: authority no longer resides securely in bounded disciplines, citation practices or expert communities, but circulates through hybrid infrastructures of platforms, databases, chatbots, search engines and institutions. The SEI framework therefore shifts attention from classification to mediation, asking how credibility is produced through infrastructural typologies, power signatures, symbolic compressions and breakdown dynamics. Its theoretical novelty lies in treating failure not as anomaly but as revelation: when infrastructures misfire, symbolic terms drift, or algorithmic systems simulate authority, the hidden scaffolding of knowledge becomes visible. As a case study, the paper’s analysis of LLMs shows how machine-generated coherence destabilises peer review, citation, authorship and scholarly temporality, producing a polytemporal zone where instant algorithmic synthesis collides with slower practices of validation. The framework also introduces a demanding ethical register, including temporal ethics, conceptual hospitality and hermeneutic patience, positioning theory as a structure designed for future transformation rather than permanent sovereignty. Ultimately, Kelly’s argument reframes knowledge organisation as the cultivation of reflexive, adaptive and justice-oriented infrastructures capable of sustaining inquiry amid instability, opacity and computational acceleration.