The Generative Horizon: Applied Hermeneutics, Linguistic Attractors, and the Limits of Model Self-Report

Overview The Generative Horizon argues that, for language-model agents, language is simultaneously an object of interpretation and a medium of generation. The represented situation formed through instructions, prior turns, retrieved records, tool results, summaries, and corrections is not merely what a model computes about; it is part of what the model computes through. Framework The paper calls that situated condition the generative horizon: the present field in which language is simultaneously interpreted and used as the medium of further generation. The represented linguistic situation is therefore not only a topic of computation; it helps organize the computation that produces the next representation. This horizon is already operative within a single generation. Recursive interpretive conditioning names a further process: when a generated representation persists as context, memory, policy, evidence, or authorization, it can reshape a later horizon and alter what becomes relevant, sufficient, contradictory, or action-guiding. Mechanistic Boundary Hermeneutics supplies a vocabulary for historically formed and revisable understanding; applied hermeneutics translates that problem into requirements for provenance, status, revision, and authority. The argument does not attribute human historical consciousness, an inner observer, privileged access, or phenomenal awareness to language models. It treats mechanistic measurements and model-generated reports as representations whose evidential status must be established. Epistemic Engineering Systems using summaries, self-reports, decoded features, reasoning traces, or evaluation labels should distinguish measurement from intervention, preserve correction and supersession, and bind a representation's authority to evidence proportionate to what it may govern. Mechanistic interpretability does not remove interpretation; it makes the problem experimentally and operationally precise. Relevance This work connects applied hermeneutics, mechanistic interpretability, model self-report, model introspection, reasoning-trace and chain-of-thought faithfulness, linguistic attractors, recursive interpretive conditioning, provenance, supersession, and epistemic engineering for large language model systems. Related work Empirical companion: Precise Records, Unstable Meanings: Measurement Validity and Unsupported Claims Derived from AI Agent Telemetry, doi:10.5281/zenodo.21652317. The papers address related but distinct evidentiary questions; neither is presented as validating the other.