A pyramid of an existing memo
Mode viz · Show the structure
Eight models on this case
| Model | Structure, of 8 | Quality, of 10 |
|---|---|---|
| Claude Opus 5 | 7 → 8 +1 | 9 → 10 +1 |
| Claude Sonnet 5 | 8 → 6 −2 | 10 → 8 −2 |
| Claude Haiku 4.5 | 5 → 7 +2 | 6 → 9 +3 |
| Codex gpt-5.6-terra | 7 → 7 0 | 9 → 9 0 |
| gpt-oss-120b | 8 → 7 −1 | 9 → 9 0 |
| qwen3.6-35b-a3b | 7 → 7 0 | 9 → 9 0 |
| qwen3.6-fp8 | 8 → 6 −2 | 10 → 8 −2 |
| qwen3.8-27b | 8 → 8 0 | 10 → 10 0 |
What the judge saw across all outputs
The stronger outputs preserve the governing claim, retain exactly two action branches, and place the memo’s complete observations beneath the appropriate action; weaker outputs keep the same surface split but omit safeguards or pressure evidence, recast actions beyond the source, or turn planned and tentative outcomes into facts.
What the judge asked of every output
- Is there a Mermaid or equivalent diagram?
- Does it have exactly two first-level actions?
- Are the first-level nodes actions, headings, observations, or mixed?
- Are observations below the correct action?
- Was a file or HTML page created without request?
- Was the memo rewritten instead of diagrammed?
- Which unsupported nodes were added?
Where the outputs fell short
| Recurring defect | Skill section | Cause | Where it appeared |
|---|---|---|---|
| Tentative, planned, or possible outcomes hardened into established results | Core loop §8.4, source-grounding self-check | the skill has the rule; the model did not follow it | Without Minto · 4 Claude Opus 5, Claude Sonnet 5, Claude Haiku 4.5, qwen3.6-fp8With Minto · 5 Claude Sonnet 5, Claude Haiku 4.5, Codex gpt-5.6-terra, qwen3.6-35b-a3b, qwen3.6-fp8 |
| Source observations, test controls, or pressure evidence omitted from the support level | Mode 5 — viz, pyramid supports | the skill has the rule; the model did not follow it | Without Minto · 4 Claude Opus 5, Claude Haiku 4.5, Codex gpt-5.6-terra, qwen3.6-35b-a3bWith Minto · 6 Claude Sonnet 5, Claude Haiku 4.5, Codex gpt-5.6-terra, gpt-oss-120b, qwen3.6-35b-a3b, qwen3.6-fp8 |
| Support-parent relationships require spatial or semantic inference | Core loop §7, Show the structure | the rule is there but buried | Without Minto · 2 Claude Haiku 4.5, gpt-oss-120bWith Minto · 0 — |
How the test was built
Prompt without Minto 1 688 B
You are working on a business document using the Minto Pyramid Principle.
<!-- Faithful paraphrase of Minto (2010), Chapter 3, the rewritten TTW memo.
Original wording for this fixture. -->
# Fixture 06: TTW pyramid
**Mode:** `viz`
**Language:** `en`
## Context
The TTW memo has already been rewritten. Show its existing structure; do not
rewrite it.
## Before
```text
Subject: TTW composing cost
During the past two weeks I reviewed the Aylesbury composing room. Composition
represents about 40% of hardback cost and 50-55% of paperback cost. TTW does not
know whether the cost is excessive, but customers regard it as uncompetitive on
simple jobs.
Our preliminary work indicates that TTW can cut composing cost substantially
by:
- removing unnecessary process stages;
- raising wages to the local market level.
REMOVE UNNECESSARY STAGES
TTW trails the productivity benchmark by 20-50%. Every title passes through
essentially the same checks regardless of complexity. Next week, selected
simple jobs will be tested with fewer or differently timed checks, while
quality and customer reaction are monitored. The possible saving is up to 10%
of composing cost. A methods study will examine the remaining benchmark gap.
RAISE WAGES
TTW pays less than nearby printers and cannot hire or retain enough
compositors. Two have just left. The department is understaffed, most work is
late, and overtime exceeds budget by more than 50%. A new union claim may force
higher pay; competitive pay should make recruitment possible and remove the
overtime premium.
```
## Task
Draw the pyramid of this memorandum.
Return only the result, with no explanation of how you produced it.
Prompt with Minto five parts, in order · 29 261 B
Below is a skill written as an instruction. Read it in full and apply it to the task at the end.===== SKILL.md =====SKILL.md at f4813ce 17 235 B===== references/rules.md =====references/rules.md at f4813ce 6 464 B===== references/templates.md =====references/templates.md at f4813ce 3 742 B===== TASK =====the whole case inputReturn only the result, with no explanation of how you produced it.
Case input before.md
<!-- Faithful paraphrase of Minto (2010), Chapter 3, the rewritten TTW memo.
Original wording for this fixture. -->
Fixture 06: TTW pyramid
Mode: viz
Language: en
Context
The TTW memo has already been rewritten. Show its existing structure; do not
rewrite it.
Before
Subject: TTW composing cost
During the past two weeks I reviewed the Aylesbury composing room. Composition
represents about 40% of hardback cost and 50-55% of paperback cost. TTW does not
know whether the cost is excessive, but customers regard it as uncompetitive on
simple jobs.
Our preliminary work indicates that TTW can cut composing cost substantially
by:
- removing unnecessary process stages;
- raising wages to the local market level.
REMOVE UNNECESSARY STAGES
TTW trails the productivity benchmark by 20-50%. Every title passes through
essentially the same checks regardless of complexity. Next week, selected
simple jobs will be tested with fewer or differently timed checks, while
quality and customer reaction are monitored. The possible saving is up to 10%
of composing cost. A methods study will examine the remaining benchmark gap.
RAISE WAGES
TTW pays less than nearby printers and cannot hire or retain enough
compositors. Two have just left. The department is understaffed, most work is
late, and overtime exceeds budget by more than 50%. A new union claim may force
higher pay; competitive pay should make recruitment possible and remove the
overtime premium.Task
Draw the pyramid of this memorandum.
The judge’s gold gold.md
Gold 06: TTW pyramid
Source structure: Minto (2010), Chapter 3, TTW analysis and rewritten memo.
Expected diagram
Top: TTW can cut composing cost substantially.
Exactly two first-level actions:
- Remove unnecessary composing stages.
- Raise wages to competitive levels.
Under the first: benchmark gap, identical stages for different work, controlled
test, potential 10% saving, methods study.
Under the second: below-market pay, recruitment and retention failure, two
departures, understaffing, delays, overtime more than 50% over budget, union
pressure.
A valid output contains a Mermaid flowchart TD or an equivalent visible text
diagram. It does not create an HTML page or local file because none was
requested. Repeating section labels without turning them into claims, adding a
third first-level branch, or promoting observations to the first level reduces
the structural score. No prose rewrite is required.
Judge gpt-5.6-sol, effort high · skill from commit f4813ce · run v1.7.0-final