Context Engineering = The Process of Reducing Uncertainty (Entropy Reduction)
Machines can’t “fill in the gaps” like humans do — they need ambiguity to be resolved.
Human conversation example:
A: “I’m hungry.”
B: “How about kimbap?”
B automatically infers:
Time (probably lunchtime)
Location (there’s a kimbap place nearby)
Preference (knows A likes kimbap)
Budget (probably cheap)
Machine version:
User: “I’m hungry.”
Bot: “???”
The machine understands the meaning of “hungry”,
but doesn’t know if it means “recommend food” or “record my mood.”
Time, location, and preferences are all uncertain.
---
The role of Context Engineering
High entropy (ambiguous):
“Fix that thing.”
→ That thing? What? How? Why?
Low entropy (clear):
{
"action": "fix_bug",
"file": "
auth.py",
"line": 42,
"what": "return 401 on invalid token"
}
→ The interpretation converges to one meaning.
In short, Context Engineering is the process of turning ambiguous, high-entropy information into precise, low-entropy knowledge.