memoryrot.com · entry no. 1

memory rot

/ˈmɛm.ər.i rɒt/ · noun

1.The progressive degradation of an AI system's contextual reliability as accumulated conversation history, retrieved documents, or self-generated content exceed what it can coherently track — manifesting as forgotten constraints, contradicted earlier statements, and silently drifting objectives.
2.informal. By extension, the broader erosion of trust in an AI-assisted workflow as unreviewed, low-quality generated output accumulates faster than it can be checked.

Etymology: Modeled on “brain rot,” applied to the specific failure modes of long-running language-model sessions.
First attested: circa 2024, in engineering discussions of long-context LLM degradation.

section i

Five symptoms

Memory rot is not a single failure but a family of them. These are the five most commonly observed and named in current usage, each cross-referenced in the glossary below.

context rot

Degradation from context overload

The progressive loss of coherence and reliability in a language model's output as the volume of context — conversation history, retrieved documents, tool outputs — grows beyond what the model weighs evenly, causing early instructions to be silently deprioritized in favor of more recent tokens.

slop debt

Unreviewed output as compounding liability

The accumulating cost of reviewing, correcting, and maintaining low-quality machine-generated content that was produced faster than it could be verified. Analogous to technical debt, but the principal is unverified prose, code, or data rather than architectural shortcuts.

instruction decay

Early instructions losing force over time

The tendency for a system prompt or early user instruction to lose influence over a model's behavior as a session lengthens, even when the instruction was never explicitly revoked.

hallucination drift

Fabrication compounding turn over turn

A gradual, session-long increase in fabricated claims that occurs as a model reasons over its own prior — possibly incorrect — outputs rather than re-grounding each turn in verified source material, so errors compound on errors.

prompt fatigue

Iterative correction overfitting the model

The diminishing quality of responses to iteratively refined prompts, observed when repeated correction attempts cause a model to overfit to the immediate correction at the expense of the original task.

section ii

Related terms

Ten terms in current circulation for describing how AI systems degrade, drift, and accumulate error over time.

context rot: The progressive loss of coherence and reliability in a language model's output as the volume of context — conversation history, retrieved documents, tool outputs — grows beyond what the model weighs evenly, causing early instructions to be silently deprioritized in favor of more recent tokens.
slop debt: The accumulating cost of reviewing, correcting, and maintaining low-quality machine-generated content that was produced faster than it could be verified. Analogous to technical debt, but the principal is unverified prose, code, or data rather than architectural shortcuts.
instruction decay: The tendency for a system prompt or early user instruction to lose influence over a model's behavior as a session lengthens, even when the instruction was never explicitly revoked.
hallucination drift: A gradual, session-long increase in fabricated claims that occurs as a model reasons over its own prior — possibly incorrect — outputs rather than re-grounding each turn in verified source material, so errors compound on errors.
prompt fatigue: The diminishing quality of responses to iteratively refined prompts, observed when repeated correction attempts cause a model to overfit to the immediate correction at the expense of the original task.
model collapse: A degenerative process in which a model trained on its own — or other models' — synthetic outputs progressively loses fidelity to the true underlying data distribution across generations of retraining.
context window: The fixed span of tokens — spanning prompt, history, and generation — that a model can attend to at once; the finite resource whose exhaustion or overextension gives rise to context rot.
synthetic data poisoning: The contamination of a training corpus with low-quality or self-referential machine-generated content, accelerating model collapse in subsequent training runs.
feedback loop pollution: The compounding effect of an AI system's outputs re-entering its own inputs — via logs, retrieved memory, or web content — without a quality filter, amplifying prior errors rather than correcting them.
recency bias (model): A model's tendency to overweight the most recently presented tokens relative to earlier ones — a mechanical contributor to both context rot and instruction decay.

section iii

New terms are added as they enter circulation. Leave your address for occasional notice of new entries — no more than a few times a year.