# Context Bloat Context bloat is the accumulation of context entries, instructions, rules, and memories without adequate pruning or prioritization. It happens when AI context is treated as append-only: new entries get added, but nothing gets removed or consolidated. The result is a context window filled with redundant, contradictory, or low-value information that dilutes the signal. This directly wastes the [[Token Budget]], since [[AI context is finite with diminishing returns]]. Past a certain point, adding more context actively degrades output quality rather than improving it. Context bloat accelerates both [[Context Drift]] and [[AI Context Rot]]. The more entries there are, the harder it becomes to maintain them, and the more likely contradictions emerge. A set of 200 rules is harder to keep consistent than 20. The antidote is [[Context Hygiene]]: regularly pruning, consolidating, and prioritizing context entries. [[Progressive Disclosure]] and [[Prompt Lazy Loading AI Design Pattern (PLL)]] help structurally by ensuring only relevant context gets loaded for a given task, rather than dumping everything into the window at once. [[Context Compression]] techniques can also reduce volume while preserving essential information. ## References - ## Related - [[Token Budget]] - [[AI context is finite with diminishing returns]] - [[Context Hygiene]] - [[Context Drift]] - [[AI Context Rot]] - [[Context Compression]] - [[Progressive Disclosure]] - [[Prompt Lazy Loading AI Design Pattern (PLL)]] - [[Context Engineering]] - [[Natural tension between compression and context]] - [[Obsidian Starter Kit - Tutorial - Managing AI sessions]] - Operational guidance to prevent context bloat in OSK sessions