Context Collapse
ELI5 — The Vibe Check
Context collapse is when the AI has been helping you perfectly for an hour and then suddenly forgets everything. It starts suggesting things that contradict what you agreed on five messages ago. It reimports libraries you removed. It rewrites code you already approved. The context window got too full or important details fell off the edge. Your productive session just turned into a frustrating loop.
Real Talk
Context collapse in AI-assisted development occurs when an LLM loses track of previously established decisions, patterns, or constraints due to context window limitations, attention degradation over long sequences, or context eviction in conversation management systems. It manifests as inconsistent suggestions, contradictory code changes, or regression to default patterns that ignore project-specific conventions.
When You'll Hear This
"Context collapse hit at 50K tokens — Claude forgot our entire API schema." / "I start a new chat every 30 minutes to avoid context collapse."
Related Terms
Context Stuffing
Context stuffing is cramming your entire codebase, README, package.
Context Window
A context window is how much text an AI can 'see' at once — its working memory.
The Loop
The Loop is that endless cycle where you tell the AI about a bug, it suggests a fix, the fix creates a new bug, you tell the AI about that bug, it suggests...
Token Anxiety
Token anxiety is that creeping dread when you're deep in a conversation with an AI and you start worrying about running out of context window.