---
license: cc-by-4.0
pretty_name: Trismegistus Receipt-Bound AI Expert Partner Evaluation
tags:
  - artificial-intelligence
  - ai-agents
  - evaluation
  - benchmarks
  - continuity
  - model-agnostic
  - research-paper
  - zenodo-doi
---

# Trismegistus: Receipt-Bound AI Expert Partner Evaluation

Author: Dean Patterson, Renaissance Field Lite

Zenodo DOI: [10.5281/zenodo.21417649](https://doi.org/10.5281/zenodo.21417649)

Zenodo record: [https://zenodo.org/records/21417649](https://zenodo.org/records/21417649)

## Summary

Trismegistus is a receipt-bound AI Expert Partner architecture developed by
Renaissance Field Lite to preserve continuity across research, coding, browser
action, source review, outreach, and paid-work operations.

The central contribution is a model-agnostic continuity and evaluation method:
run a baseline route, run an architecture-on route, evaluate the same task under
the same source and evidence gate, and preserve SQL, JSON, Markdown, media,
benchmark, source, or browser receipts for review.

This Hub package mirrors the DOI-backed public-safe paper bundle for AI-native
discovery. The official archival publication is the Zenodo record above.

## Paper Files

- [`files/TRISMEGISTUS_RECEIPT_BOUND_AI_EXPERT_PARTNER_EVALUATION_ARXIV_CANDIDATE_2026-07-16.pdf`](files/TRISMEGISTUS_RECEIPT_BOUND_AI_EXPERT_PARTNER_EVALUATION_ARXIV_CANDIDATE_2026-07-16.pdf)
- [`files/TRISMEGISTUS_WHITE_PAPER_PUBLIC_SAFE_2026-07-15.md`](files/TRISMEGISTUS_WHITE_PAPER_PUBLIC_SAFE_2026-07-15.md)

## Figures

- [`figures/figure_01_three_surface_architecture.png`](figures/figure_01_three_surface_architecture.png)
- [`figures/figure_02_receipt_bound_operator_loop.png`](figures/figure_02_receipt_bound_operator_loop.png)
- [`figures/figure_03_baseline_vs_architecture_on.png`](figures/figure_03_baseline_vs_architecture_on.png)
- [`figures/figure_04_six_lane_field_curriculum.png`](figures/figure_04_six_lane_field_curriculum.png)

## Public-Safe Boundary

This package publishes public-safe paper materials only. It does not include
private raw logs, credentials, browser profiles, local databases, biometric
captures, unpublished security artifacts, or private operator-state material.

The paper does not claim a completed AGI system, official public benchmark
leaderboard result, clinical system, or peer-reviewed status. It claims a
receipt-bound engineering method and identifies the current evidence, boundary,
and next gate.

## Citation

```bibtex
@misc{patterson2026trismegistus,
  author = {Patterson, Dean},
  title = {Receipt-Bound AI Expert Partner Evaluation: A Model-Agnostic Continuity Architecture for Baseline Versus Architecture-On Workflows},
  year = {2026},
  publisher = {Zenodo},
  doi = {10.5281/zenodo.21417649},
  url = {https://doi.org/10.5281/zenodo.21417649}
}
```