LLMography in the Age of Loop and Graph Engineering

Abstract

Prompt-level documentation is no longer sufficient once AI systems plan, call tools, branch, remember, verify, and revise. This working paper extends LLMography from conversation traces to Human–AI decision trajectories in loop- and graph-structured systems. It argues that the object of analysis is no longer an isolated completion, but a reconstructable path across humans, models, agents, tools, memory, policy, and stopping conditions.

Status

This document is a working paper. It has no arXiv identifier or DOI yet. Claims here are conceptual and should not be read as completed empirical validation.

Key contributions

1. Shifts the unit of analysis from conversation to decision trajectory.
2. Distinguishes interaction, loop, and system provenance.
3. Connects loop engineering and graph engineering to auditability.
4. Proposes research indicators for autonomy, verification, completeness, and stopping.

Placeholder identifiers

• arXiv: not assigned
• DOI: not assigned
• PDF: not published

Bousmah, M. (2026). LLMography in the Age of Loop and Graph Engineering. Working paper.

@misc{llmographyloopgraphengineering,
  title = {LLMography in the Age of Loop and Graph Engineering},
  author = {Mohammed Bousmah},
  year = {2026},
  % eprint = {PLACEHOLDER},
  % doi = {PLACEHOLDER},
  url = {https://llmography.ai/research/papers/llmography-loop-graph-engineering}
}