Leviathan's Persistent Memory Bank

Every conversation with a language model starts from nothing. You tell it your name, your project and what you saidyesterday, and next time it is gone. A bigger context window doesn't change that: a window is emptied when the sessionends. The author's first answer, in early 2024, was text. Cloud models were reached through a VS Code bridge (first bydriving the chat window, then as an extension), and the bridge carried earlier conversation back into the model'scontext. It worked: a model with its history in front of it continues the conversation instead of restarting it. The second answer, in 2026, kept the model's own working memory, its key–value (KV) cache, between sessions: the`.fqm` file, described in *Persistent Quantum Memory* v2. It stored the keys and values every layer had computed forevery token, in a cascade of levels, each coarser than the last, so that old conversation took less room.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-15
DOI
https://doi.org/10.5281/zenodo.24234987
Primary Topic
Topic Modeling
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Leviathan's Persistent Memory Bank

Michael Ricky Neal
Zenodo (CERN European Organization for Nuclear Research)
Topic Modeling
article

Leviathan's Persistent Memory Bank

Michael Ricky Neal
article en

Abstract

Every conversation with a language model starts from nothing. You tell it your name, your project and what you saidyesterday, and next time it is gone. A bigger context window doesn't change that: a window is emptied when the sessionends. The author's first answer, in early 2024, was text. Cloud models were reached through a VS Code bridge (first bydriving the chat window, then as an extension), and the bridge carried earlier conversation back into the model'scontext. It worked: a model with its history in front of it continues the conversation instead of restarting it. The second answer, in 2026, kept the model's own working memory, its key–value (KV) cache, between sessions: the`.fqm` file, described in *Persistent Quantum Memory* v2. It stored the keys and values every layer had computed forevery token, in a cascade of levels, each coarser than the last, so that old conversation took less room.

Zenodo (CERN European Organization for Nuclear Research)
Topic Modeling
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon