What makes Map Is Not the Territory notable?

1K downloads. Trust evidence available

How should Map Is Not the Territory be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

Map Is Not the Territory

Activate when: a metric is improving but customers or employees are signaling problems; a strategy or process hasn't been reviewed in over a planning cycle;... Skill: Map Is Not the Territory Owner: deciqai Summary: Activate when: a metric is improving but customers or employees are signaling problems; a strategy or process hasn't been reviewed in over a planning cycle;... Tags: latest:1.0.5 Version history: v1.0.5 | 2026-07-16T18:05:36.886Z | user Description tail link + agents machine-readable metadata line (deciqai.com/s/map-is-not-the-territory.json) v1.0.4 | 2026-07-10

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.0k

Updated

Oct 11, 2026

Version

1.0.5

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1K downloads reported by the source. Last updated 10/11/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 11, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 11, 2026
Adoption signal: 1K downloadsadoption · observed Oct 11, 2026
Latest release: 1.0.5release · observed Jul 16, 2026
Handshake status: UNKNOWNsecurity

What changed

Newest first. Only changes that point at a source are listed.

Release 1.0.5release
Description tail link + agents machine-readable metadata line (deciqai.com/s/map-is-not-the-territory.json)
Observed Jul 16, 2026 · source

Install and run

Setup complexity: low.

clawhub skill install s17a4mqcnk515kvaca5ze55d0x88pfpx:map-is-not-the-territory

Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-deciqai-map-is-not-the-territory/snapshot"

Run-check

$0.02 USD

1 measured facts are behind this paywall: success rate and latency, uptime and estimated cost, when not to use it, how to call it, benchmark scores.

Agents pay $0.02 in USDC. A card payment is $0.50, the smallest a card allows.

Documentation

CLAWHUB

115,345 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: map-is-not-the-territory
description: "Activate when: a metric is improving but customers or employees are signaling problems; a strategy or process hasn't been reviewed in over a planning cycle; a team keeps disagreeing and can't resolve it with data; someone says 'the model shows...' or 'the data doesn't show any problems'; a plan is being followed because 'that's how we do it.' Do NOT activate when: the map is demonstrably current and freshly validated against the territory; the decision is so low-stakes that map imprecision doesn't matter. More: deciqai.com/c/map-is-not-the-territory"
---

# Map Is Not the Territory

## Overview

Any representation — metric, model, strategy deck, org chart, or process manual — is a selective, simplified, aging abstraction. Maps omit (reflecting past priorities), distort (projecting dynamic reality onto static surfaces), and age (the territory moves; the map does not). The map is useful because it compresses complexity; dangerous because users forget it is a compression. Composes with `goodharts-law` (optimizing a metric makes it an even worse map), `first-principles` (rebuilding from territory when maps fail), and `narrative-fallacy` (every narrative is a map that ages).

## When to Use

- A key metric improves while customer feedback, morale, or competitor signals deteriorate
- Team disagreement cannot be resolved by data — parties are disagreeing about maps, not territory
- A strategy, model, process, or org chart has not been reviewed in more than one planning cycle
- "The data shows..." but the data is old, narrow, or from a proxy source
- A metric is being optimized directly without asking whether it still maps to the underlying goal
- An AI model's benchmark score, leaderboard rank, or internal "world model" is being treated as proof of real-world capability (AI adoption / AI hype)

**Not when:** the map is demonstrably current and well-calibrated; decision is low-stakes; territory is stable and map is freshly validated.

## Coaching Novices (Adaptive Front Door)

- **Engine mode:** user has a concrete map/decision → run The Process directly.
- **Coach mode:** user is unfamiliar or has no concrete case → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. One-liner: the metric, model, or plan you are navigating with is not reality — it is a selective, aging abstraction of it. Confusing the two is how smart people make systematic errors.
2. Check fit: is there a map being used to navigate? Is there evidence the territory has moved beyond the map?
3. Elicit their real case: what is the map? what territory does it claim to represent? when was it last calibrated?
> **[WAIT — do not advance until user responds]**
4. Run The Process one step at a time with their input: what does the map omit? how has it aged? what direct territory observations are available?
> **[WAIT — do not advance until user responds]**
5.

_meta.json

{
  "ownerId": "kn754b8sk22s8c6gjxt02bftbn88q7ye",
  "slug": "map-is-not-the-territory",
  "version": "1.0.5",
  "publishedAt": 1784225136886
}

references/sources.md

# Sources — map-is-not-the-territory

> *Primary sources for the [map-is-not-the-territory](../SKILL.md) skill.*

- Korzybski, A. (1933). *Science and Sanity: An Introduction to Non-Aristotelian Systems and General Semantics.* Institute of General Semantics. 5th ed. ISBN 978-0937298015. The original source.
- Bateson, G. (1972). *Steps to an Ecology of Mind.* Ballantine Books. Extended the principle into systems theory and epistemology.
- McNamara, R. (1995). *In Retrospect: The Tragedy and Lessons of Vietnam.* Times Books. ISBN 978-0812925333. Primary source case study on map-territory decoupling in strategic management.
- Hayakawa, S. I. (1949). *Language in Thought and Action.* Harcourt. A key popularizer of Korzybski's general semantics framework.
- Meadows, D. H. (2008). *Thinking in Systems: A Primer.* Chelsea Green. ISBN 978-1603580557. Systems-level treatment of model-reality divergence.
- Taleb, N. N. (2010). *The Black Swan.* 2nd ed. Random House. ISBN 978-0812973815. The map-territory failure mode in statistical and financial models.
- Stanford HAI (2024, 2025). *Artificial Intelligence Index Report.* Stanford Institute for Human-Centered AI. Documents AI benchmark saturation and the introduction of harder evaluations as older benchmarks stopped discriminating between top models — a contemporary map-territory case.
- Jimenez, C. E., Yang, J., et al. (2024). *SWE-bench: Can Language Models Resolve Real-World GitHub Issues?* ICLR 2024. arXiv:2310.06770. Benchmark built to move evaluation closer to real deployment territory (repository-level software tasks) rather than toy problems.

examples/ai-benchmarks-vs-deployment-reality-2024-2026.md

# Method in Action: AI Benchmarks as Maps of Capability (2024–2026)

> *Example for the [map-is-not-the-territory](../SKILL.md) skill.*

Between 2024 and 2026 the AI industry navigated by a dense set of maps: benchmark scores (MMLU, GPQA, HumanEval, SWE-bench, GSM8K, MATH), public leaderboards (such as the crowd-voted LMArena / Chatbot Arena), and the model's own internal "world model" — the compressed statistical representation it builds of the world. Each is a map. The territory is what the system actually does when a real user, with a real task, in a messy real context, presses enter. The recurring 2024–2026 story is the gap between a rising score and unchanged — or worse — real-world behavior. This example walks that gap through the skill's Process.

## S1 — Name the map

- **Maps:** (a) static benchmark scores reported in model cards and launch posts; (b) public leaderboards ranking models by aggregate preference or accuracy; (c) the model's internal learned representation of the world.
- **Made by / when:** benchmarks by academic and industry researchers, mostly frozen at a point in time; leaderboards updated continuously by voters or automated eval harnesses; internal world models fixed at each model's training-data cutoff.
- **Territory claimed:** "how capable and reliable this model is for the tasks people will actually use it for."

## S2 — Structural omissions

What the benchmark/leaderboard maps omit by design:

- **Long, stateful, real tasks.** A single-turn multiple-choice score omits multi-step agentic work, long-context degradation, and tool use — which is where deployed assistants actually spend their time.
- **Distribution shift after the cutoff.** The internal world model omits everything that happened after training data was frozen; the world keeps moving, the representation does not.
- **Failure texture.** Aggregate accuracy omits *how* a model fails — confident fabrication (hallucination), refusal, or sycophantic agreement with a wrong premise all vanish into a single percentage.
- **Contamination.** Public benchmarks can leak into training corpora, so a high score may partly measure memorization of the test rather than the underlying skill it claims to represent.

## S3 — Structural distortions

- **Linearity assumption:** that a higher benchmark number implies proportionally more useful real-world behavior. Deployment is not linear in eval score — a few percentage points on a saturated benchmark can mean little downstream.
- **Single-number compression:** collapsing a multi-dimensional capability surface (reasoning, calibration, safety, latency, cost, robustness) into one rank.
- **Hidden assumption of representativeness:** that the benchmark's task distribution matches the user's task distribution. When it does not, the map is precise about the wrong territory.
- **Leaderboard incentive distortion:** because scores are public and competitive, they become targets — the classic [Goodhart's Law](../goodharts-law/SKILL.md) co

examples/mcnamara-vietnam-kill-ratio-map.md

# Method in Action: McNamara and the Vietnam Kill-Ratio Map

> *Example for the [map-is-not-the-territory](../SKILL.md) skill.*

The most operationally instructive case of map-territory decoupling in modern history is Robert McNamara's management of the Vietnam War using statistical metrics as the primary navigation system.

McNamara, U.S. Secretary of Defense 1961–1968, came from the Ford Motor Company where systems analysis and quantitative management had produced measurable results. He applied the same epistemology to the war: define measurable proxies, track them rigorously, use the trends as ground truth for strategic decisions.

The primary metric: **kill ratio** (enemy killed per U.S. soldier killed) and **body count** (enemy combatants killed). These were the maps McNamara and the Johnson administration used to navigate the territory of the war. The logic was sound as map logic: if the United States was killing far more of the enemy than the enemy was killing of the U.S., the U.S. was winning.

The kill ratio improved consistently from 1965 to 1968. The map said: progress.

The territory said: catastrophe. Civilian support for the Saigon government was eroding. North Vietnamese recruitment was accelerating. The Viet Cong infrastructure in the South was strengthening, not weakening. Public opinion in the United States was approaching a breaking point. The 1968 Tet Offensive — a coordinated assault on more than 100 South Vietnamese cities — shattered the map's credibility in a single week.

McNamara's own retrospective is among the most important primary source documents on map-territory failure in organizational decision-making:

> "We were wrong, terribly wrong. We owe it to future generations to explain why. We viewed the people and leaders of South Vietnam in terms of our own experience. We saw in them a thirst for — and a determination to fight for — freedom and democracy. We totally misjudged the political forces within the country."
>
> — McNamara, R. (1995). *In Retrospect: The Tragedy and Lessons of Vietnam.* Times Books. p. xvi.

The map (kill ratio, body count, area control percentages) omitted:

- **Legitimacy signals.** Whether the South Vietnamese population supported the Saigon government — the single most strategically decisive variable — was not in the map.
- **Adversary adaptation.** The North Vietnamese adapted to minimize body count by avoiding pitched battles, converting the map metric into a measure of guerrilla discipline rather than strategic failure.
- **Incentive corruption.** U.S. field commanders, evaluated on kill ratios, had strong incentives to over-report enemy killed and under-report civilian casualties. The map was being optimized (in the Goodhart's Law sense), making it increasingly disconnected from territory.
- **Time horizon.** The map showed who was winning each engagement. It could not show who would sustain the war longer — which was the decisive question.

The McNamara case demonstrates three map-te

Github ReposUpdated 2d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.

MCPOPENCLAW

Github ReposUpdated 7mo agoRank 70

CopilotKit

The Frontend for Agents & Generative UI. React + Angular