moizIndependent mind.
Engineer by practice.
Résumé
Menu

Independent research — Abdul Moiz Ahmed

Hashing
is not
retrieval.

A three-arm study of context accumulation in small-model code cartography.

Read the paper · archived on Zenodo

A smaller history.
A clearer handoff.

When one agent hands work to another, what should travel with it? The entire conversation, a compressed identifier, or a retrievable artifact?

This study isolates how those choices change the context sent to workers. Structured artifact passing reduced peak context by 38.3% compared with full-history passing in the controlled evaluation.

This is a result from a specific two-worker code-cartography harness. It does not establish a general improvement in reasoning quality.

Pass the work three ways.

One orchestrator hands a job to a worker. The only thing that changes is what travels with it.

What crosses the wire
TaskTurn 1Turn 2Turn 3Turn 4Turn 5Turn 6

Every previous turn, verbatim

Peak context sent to the worker
100%
Baseline

The record survives intact, and so does everything the worker did not need. Context grows with the conversation rather than with the task.

Normalized to the full-history baseline, measured inside a two-worker code-cartography harness. A result about context cost in that apparatus — not a general claim about reasoning quality.

Three ways to pass the work.

Full history

Pass the accumulated conversation forward. It preserves the record, but grows the amount of context workers receive.

Opaque compression

Pass a compressed representation. A short identifier alone cannot make the underlying state recoverable.

Structured artifacts

Pass inspectable outputs with explicit retrieval. Measure the context actually sent to the worker, then compare it with the baseline.

What this
proves. And
what it doesn't.

  • A three-arm design with a preserved, re-runnable apparatus
  • A 38.3% peak-context reduction

Filed,
not just
hosted.

The study is archived on Zenodo, the open research repository operated by CERN, under a DOI that keeps resolving no matter what happens to this site. What is filed there is the apparatus, not a summary of it.

DOI
10.5281/zenodo.22844754 · resolves to the latest version
Filed
5 September 2026
Version
1.0.2 · 19 September 2026 · 10.5281/zenodo.22848744
Contents
The preprint, and a reproducibility bundle: the executable apparatus, frozen results, task fixtures, telemetry and verification instructions.
Licence
CC BY 4.0. Cite it, reuse it, re-run it.
Affiliation
FAST National University of Computer and Emerging Sciences
Open the archived recordRead the paper · PDF

A DOI makes the record permanent and citable. It is not peer review, and it does not make the result any more general than the harness it was measured in.

What I'd
test next.

Longer agent chains and broader task classes. The useful question is how well the handoff contract survives different workloads, and where its costs start to show.

Inspect the full study · DOI record