Concept

Context window

You might say: context budget, working context

Everything the model sees on each model provider request. It is finite and is the only surface through which the model perceives.

What it does
Carries a run's working state from turn to turn. A fresh run can start with an empty window, so sizing a task to one session keeps it in the sharpest part of the session.
Ask
What must this run see now, and what can it leave behind?
Trouble
Quality drops as the window fills. Attention degrades, and compaction or autocompact can lose detail; clearing discards the working state.

What to look for

  1. A context meter fills as messages, model responses, and tool results accumulate
  2. A compaction or autocompact event replaces history with a summary
  3. A new session opens with an empty context window

Connections

Sources

  1. Matt Pocock, AI Coding Dictionary, Context windowEverything the model sees on each model provider request. Finite, model-specific, and the only surface through which the model perceives anything.
  2. Matt Pocock, AI Coding Dictionary, Context windowFinite means it fills up. Every turn appends more — your messages, the model's responses, tool results — and a long session will eventually hit the limit, forcing compaction or clearing.
  3. Matt Pocock, AI Coding Dictionary, ClearingThe next message begins with an empty session and an empty context window.
  4. Matt Pocock, AI Coding Dictionary, Smart zoneDoing one task per session gives each task the sharpest part of the session.
  5. Matt Pocock, AI Coding Dictionary, Smart zoneEarly in a session the agent is sharp and focused. As the session grows it drifts into a dumb zone: sloppier, forgetful, more mistakes.
  6. Matt Pocock, AI Coding Dictionary, Attention degradationAs a session grows, each token's attention budget is spread across more competitors. The signal on any one meaningful relationship shrinks; noise from irrelevant context crowds in.
  7. Matt Pocock, AI Coding Dictionary, CompactionWhatever didn't make it into the summary is gone from the context.
  8. Matt Pocock, AI Coding Dictionary, AutocompactCompaction is lossy, and autocompact is lossy at a moment you didn't choose.
  9. Matt Pocock, AI Coding Dictionary, ClearingWhat's gone is the agent's working state: the model is stateless, so the new session knows nothing the old one knew.

Use it on your project

teardown-apply: hold "Context window" (context-window) from the teardown at https://teardownatlas.com/software-factory/ against my project

Paste it into a coding agent that has the teardown skills.

An AI agent wrote this map with the teardown skill. It can contain errors: check the sources before you rely on a part. About and legal