Reflective Time Model (RTM), by Eran Harpaz. The Reflective Time Model research hub: documented architectures, case books, manuscripts, Null models, observer/role replacement tests, code and source material for auditing what RTM actually shows.
The findings hub presents four main areas: The Living Subscriber, Albania — Plaza Tirana, 14–15 July 2026, and a cross-case Historical Events series spanning Titanic, 9/11 and Ford Capri. Only after the cases are shown does the site compare recurring patterns. Observer replacement, role permutation and Null models remain a separate validation layer.
A foundational RTM record on Zenodo documenting an earlier stage of the framework, methodology and supporting material. It remains part of the research lineage; the current RTM synthesis is broader and is also documented across later manuscripts, case books, validation packages and public research records.
Eran Harpaz’s ORCID profile and researcher identifier: 0009-0004-3839-6124.
RTM establishes the existence of precise information architectures in which record, event, time and observer identity reconstruct one another. In the strongest bounded cases this specificity survives observer replacement, role permutation, dependence compression, knockouts, exhaustive finite assignment and Null models. The model-level description is Observer–Record Coupling: a measurable relation between observer state and the state of accessible records. Selecting the exact physical implementation is a further research layer.
9/11 — Two distinct scoring catalogues are kept separate. In the expanded catalogue the documented observer completes 365/365 versus a 59/365 maximum among 10,000,000 random observers. In the conservative linear catalogue the documented observer completes 324/324; the unoptimized random-observer maximum is 174/324 among 20,000,000 profiles, while a purpose-built genetic optimizer reaches 320/324 after 17,580,000 evaluations. These comparator families are reported separately rather than pooled.
Titanic — Across the exhaustive 11! = 39,916,800 role assignments, the documented assignment is the sole global maximum: 190/190 versus 183/190 for the best alternative assignment, and 89/89 versus 82/89 on the strict layer. Separately, 20,000,000 broad random observers reach maxima of 68/190 and 20/89. The exhaustive assignment search and the random-observer test are different comparison objects.
Emily Dickinson — The primary object completes 57/57 relations, 15/15 logical classes and 9/9 dependency blocks. Among 423,360 coherent profiles only the documented profile completes all three levels; 100,000,000 context-fixed identity-replacement worlds produce no joint full reproduction. After removing 43/57 direct observer-target relations, the remaining 14 indirect relations still reconstruct the same profile uniquely.
World Cup live-time + Spain — The complete concordance passes a deterministic 99/99 closure-unit audit. A separate frozen role-sensitive object contains 61/61 constraints compressed to 13/13 dependency blocks. Exhaustive enumeration of the legal role space finds only 2 states among 34,951,113,757,163,520,000 that complete the full architecture; they differ only by the documented F02 6↔21 symmetry and therefore form one structural solution class. The Spain source was frozen before the later outcome, while the 344 decoding remained retrospective rather than a blinded pre-outcome prediction.
Every result belongs to its declared score object, reference model and comparison space. Closure counts, role-sensitive constraints, dependency blocks, role assignments and Null worlds are not interchangeable and are not multiplied into a universal p-value. The purpose is to test whether the architecture preserves source, target, semantic role, dependency and chronology under competitive alternatives.
Source scale: does the same architecture persist when the source is a large archive, a single object, one message or a live record?
Chronology: can a frozen earlier state and later-revealed information reconstruct one another under a documented time order?
Observer specificity: does the structure remain tied to the correct identity-and-role registry when observers or roles are replaced?
Reconstruction depth: does later information return only obvious targets, or also internal nodes, representations, loops and earlier substructures?
Record state: can measurement time, state changes and the transition between record states become part of the same architecture?
Null robustness: how often do matched alternative worlds produce an architecture at least as strong under the same search rules?
After the architecture is measured, RTM proposes a general mechanism in which observer state, records, events and access times are parts of one information structure extending across the system's history. Earlier and later states can belong to the same globally consistent solution. This is a proposed mechanism, not a final identification of the physical implementation.
The official GitHub profile for RTM code, validation repositories and reproducibility materials associated with Eran Harpaz.
RTM videos, finding documentation and explanatory material.
The official creator search collects all records listed under Eran Harpaz, with direct links to selected records below.
An archived manuscript version retained for documentation, transparency and research history.
Direct link to the RTM book series on Amazon. The books section is kept separate from Zenodo research records so the two publication types remain clearly distinguished.
