Interactive lesson~18 minAdvanced
Long Context & Memory
Long-context systems decide what belongs in the prompt, what belongs in memory, and what should be retrieved only when needed.
Long contextMemoryRetrieval
Mental model
Context is working memory; retrieval is the bookshelf.
Million-token contexts are useful, but cost, attention dilution, and stale memory still require careful architecture.
Recall
balanced74% modeled signal
Cost
balanced67% modeled signal
Freshness
balanced65% modeled signal
Concept pipeline
Build the idea in four moves
Interactive lab
Design memory for a long-running assistant.
Segment
Break experience into retrievable chunks and events.
Focus lens
The part that usually clicks late
Context budget
More context is not always more attention.
Recall
74
Cost
67
Freshness
65
Knowledge check
Why not put every memory into the prompt?
Next horizon
Where this topic is headed
Episodic agent memory
Memory evals
Context compression models