# 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