Agentic AI Notebook
Context Engineering
Phase 6Module 14 of 15

Context Freshness

Agents quote last month's price with confidence.

Milk in the fridge. Check the date before you pour.

Visual Workflows

Start here — scroll inside each diagram frame to explore, then use + / to zoom up to 200% if needed.

Overview

100%
Loading diagram...

Scroll inside the frame to explore · use + / − to zoom up to 200%

100%
Loading diagram...

Scroll inside the frame to explore · use + / − to zoom up to 200%

This turn

100%
Loading diagram...

Scroll inside the frame to explore · use + / − to zoom up to 200%

Milk in the fridge. Check the date before you pour.

Key Takeaways

  • 1.Stale facts in the pack are worse than missing facts.
  • 2.TTL, updated_at, and source version belong on every memory and chunk.
  • 3.Prefer a tool refresh over a three-week-old embedding.
  • 4.Say when a fact might be stale.
  • 5.Agents quote last month's price with confidence.

Real Example

Scenario

Freshness: pack only what this turn needs.

What you would do

In Context Engineering, apply Context Freshness to this scenario: Freshness: pack only what this turn needs. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Commands

Commands to Remember

  • Stale facts in the pack are worse than missing facts
  • TTL, updated_at, and source version belong on every memory and chunk
  • Prefer a tool refresh over a three-week-old embedding
  • Say when a fact might be stale

Cheat Sheet

Quick recap

quick ref
  • Freshness
  • Budget
  • Don't dump

Common Mistakes

  • Skipping evaluation for Context Freshness before production
  • No logging or tracing around context freshness steps
  • Ignoring cost and latency implications