wiki / concepts / smart-zone
Smart Zone
Machine ingest — raw context
loading…
~… tokensappend .md to any wiki URL for this view
Smart Zone
The Smart Zone refers to the initial token budget of a session (typically between 0 and 150k tokens on modern frontier models) during which an AI agent operates at maximum reasoning sharpness, exactness, and adherence to complex instructions.
Why the Smart Zone Exists
Although modern models advertise 1M+ or 2M+ token context windows, their effective reasoning capacity is not constant throughout the entire window.
As a conversation approaches 100k–150k tokens:
- Latency and compute costs scale non-linearly.
- Subtle constraints in the system prompt are more frequently forgotten.
- Refactoring operations become sloppy and prone to regressions.
Operating in the Smart Zone
- Bounded Phases: Keep each phase (grilling, spec, implementation) within the smart zone.
- Phase Boundary Compaction: At the boundary of a phase (e.g. going from specification to implementation), compress state into a handoff artifact and start a fresh session context.
- One Ticket Per Fresh Context: Run each implementation ticket in a clean context window rather than daisy-chaining multiple major feature builds in one marathon session.
Related
context rot, handoff artifacts, context engineering, subagents and context management, ai coding taxonomy.
Evidence — verified primary sources
| aihero-dev-homepage-2026 | https://www.aihero.dev/ | ingested 2026-08-22 sha256:231157fed45f… |
| aihero-ai-coding-dictionary-2026 | https://www.aihero.dev/ai-coding-dictionary | ingested 2026-08-27 sha256:52b0a5da7c9f… |
Graph context
References (2)
Handoff Artifactsand start a fresh session context.Context Rot, handoff-artifacts, context-engineering, subagents-and-context-management, ai-coding-taxonomy. Referenced by (7)
Agent Harness EngineeringAI Coding Taxonomy & Agent Experienceearly in a session versus the sloppier "dumb zone" later in extended sessions.Codex Harness Architectureagainst context-rot.Context Rot, handoff-artifacts, context-engineering, prompt-bloat, subagents-and-context-management, ai-coding-taxonomy.Five Debts of Agentic Engineering) and inducing context-rot.Handoff Artifacts, context-rot, idea-to-ship-flow, progressive-disclosure, ai-coding-taxonomy.Think Tooland context staging: put the right information in the right place in history for the decision that needs it.