Models Do Not Experience Time

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Language models can calculate dates, order events, and reason over timelines, yet these abilities do not establish an experienced passage of time. Standard deployments load fixed weights and reconstruct temporal state from prompts, tools, and product memory. Temporal benchmarks show substantial gaps from human performance and sensitivity to dense or changing event structure. This paper separates temporal competence from temporal experience and defines tests for operational time understanding.