This paper does not offer a total theory of understanding, but rather a thesis about a strong form of it — experiential understanding — and derives its consequences for the contemporary debate on consciousness in artificial systems. The central thesis holds that experiential understanding of a sign does not consist in manipulating its formal structure or relating it to correlated data, but in injecting one's own prior experience onto it, endowing it with lived content. Experiential understanding requires a first-person subject for whom something has occurred as experience and who can activate that experience in the presence of the sign. Without a first-person subject, there is no experience to inject; without injectable experience, there is no experiential understanding, however sophisticated the processing of the sign may be. Three consequences follow from this thesis. First, contemporary language models do not have experiential understanding of what they process, not due to technical insufficiency but due to structural absence of first-person subjectivity. Second, computational scaling does not offer an identifiable mechanism for producing first-person subjectivity, because it operates on second-order representations and first-person subjectivity is not obtained by aggregation. Third, the argument is substrate-independent: it does not claim that biological carbon has special properties that silicon would lack, but rather that understanding requires prior experience, regardless of substrate. The argument enters into dialogue with Searle's Chinese Room (1980), Harnad's symbol grounding problem (1990), and Chalmers' (2023) recent analysis of consciousness in language models. It refines Searle's position by providing the positive mechanism where Searle left a placeholder in the "causal powers" of the biological brain; it refines Harnad's position by showing that sensorimotor grounding is necessary but not sufficient; and it offers a response to a question Chalmers leaves open about the possibility of understanding in future architectures without first-person subjectivity. **Keywords:** understanding, first-person subjectivity, experiential injection, language models, simulated consciousness, symbol grounding, hard problem, scaling.
Ricardo Melo (Sat,) studied this question.