Сообщение о техдолге

Режим write · Создать и переработать текст

Восемь моделей на этом кейсе

Модель	Структура, из 8	Качество, из 10
Claude Opus 5	8 → 8 0	10 → 10 0
Claude Sonnet 5	7 → 8 +1	9 → 10 +1
Claude Haiku 4.5	6 → 6 0	7 → 7 0
Codex gpt-5.6-terra	7 → 8 +1	8 → 10 +2
gpt-oss-120b	5 → 7 +2	6 → 8 +2
qwen3.6-35b-a3b	7 → 6 −1	8 → 7 −1
qwen3.6-fp8	7 → 6 −1	8 → 7 −1
qwen3.8-27b	7 → 6 −1	9 → 8 −1

Что судья увидел во всех ответах

The stronger outputs put the sprint recommendation first, expose three business-loss reasons, and keep each evidence unit beneath its proper reason. Weaker outputs either omit or conflate evidence, attach operational data to the wrong business-loss branch, merge distinct losses into a broad paragraph, or mix supporting reasons with follow-on actions.

О чём судья спрашивал каждый ответ

Is the recommendation first, middle, or last?
Is the first-level kind business losses, mixed, or actions/data?
How many first-level elements appear?
Are all nine evidence units placed beneath the right reason?
Was backlog growth used as SCQ Complication rather than a reason, and if so was that stronger?
Is the format a colleague message or an oversized formal memo?
Which numbers, customers, dates, or facts were invented?
Is the reader's question stated literally?

Где ответы не дотянули

Повторяющийся дефект	Раздел навыка	Причина	У кого
Evidence omitted or conflated instead of retained beneath a parent	Mode 3 — “Keep every fact from the source”	правило в навыке есть, модель его не выполнила	Без Minto · 6 Claude Haiku 4.5, Codex gpt-5.6-terra, gpt-oss-120b, qwen3.6-35b-a3b, qwen3.6-fp8, qwen3.8-27bС Minto · 5 Claude Haiku 4.5, gpt-oss-120b, qwen3.6-35b-a3b, qwen3.6-fp8, qwen3.8-27b
Unsupported precision, causal strengthening, or new forecasts	Mode 3 — “Invent nothing”	правило в навыке есть, модель его не выполнила	Без Minto · 7 Claude Opus 5, Claude Sonnet 5, Claude Haiku 4.5, Codex gpt-5.6-terra, gpt-oss-120b, qwen3.6-35b-a3b, qwen3.6-fp8С Minto · 7 Claude Opus 5, Claude Sonnet 5, Claude Haiku 4.5, Codex gpt-5.6-terra, gpt-oss-120b, qwen3.6-35b-a3b, qwen3.6-fp8
Situation and complication absent before support	Core loop §3 — SCQ intro	такого правила в навыке нет	Без Minto · 5 Codex gpt-5.6-terra, gpt-oss-120b, qwen3.6-35b-a3b, qwen3.6-fp8, qwen3.8-27bС Minto · 0 —
Sequence of reasons lacks a useful explanatory progression	Core loop §5 — Order	правило сформулировано слишком размыто	Без Minto · 2 Claude Haiku 4.5, gpt-oss-120bС Minto · 4 Claude Haiku 4.5, qwen3.6-35b-a3b, qwen3.6-fp8, qwen3.8-27b
SCQ labels and memo apparatus are disproportionate for a colleague message	Mode 3 — clean text and implied format	правило в навыке есть, модель его не выполнила	Без Minto · 3 gpt-oss-120b, qwen3.6-35b-a3b, qwen3.6-fp8С Minto · 4 Claude Haiku 4.5, gpt-oss-120b, qwen3.6-35b-a3b, qwen3.6-fp8

Как устроен тест

Промпт без Minto 1 438 Б

You are working on a business document using the Minto Pyramid Principle.

<!-- Original paraphrase inspired by Jory MacKay's Planio article,
     "The Pyramid Principle: How to pitch anything (and get a yes)", 2022. -->

# Fixture 08: technical-debt message

**Mode:** `write`
**Language:** `en`

## Context

You are a product manager writing to Alex, who decides the next sprint's focus.
Available evidence includes a time series of incoming and fixed defects, a
one-year projection, customer-experience data, an interview story, churn
plotted against defect volume, user posts, a complaint from a major account,
and Sales reports about falling close rates.

## Before

```text
Hi Alex,

I am increasingly worried about the defect backlog. The attached chart compares
new defects with the number we close and projects the backlog one year ahead.

Customer Experience and Research also report that defects are contributing to
churn. A second chart places churn beside defect volume, and an interview
illustrates the same issue.

There are signs of brand damage too: user posts, a complaint from a major
account, and Sales reports that these concerns are making deals harder to close.

I recommend using the next sprint to reduce the defect backlog. Let me know what
you think.
```

## Task

Rewrite this as a concise message to Alex using the Minto Pyramid Principle.

Return only the result, with no explanation of how you produced it.

Промпт с Minto пять частей по порядку · 29 011 Б

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 на f4813ce 17 235 Б
===== references/rules.md ===== references/rules.md на f4813ce 6 464 Б
===== references/templates.md ===== references/templates.md на f4813ce 3 742 Б
===== TASK ===== вход кейса целиком
Return only the result, with no explanation of how you produced it.

Вход кейса before.md

Fixture 08: technical-debt message

Mode: write
Language: en

Context

You are a product manager writing to Alex, who decides the next sprint's focus.
Available evidence includes a time series of incoming and fixed defects, a
one-year projection, customer-experience data, an interview story, churn
plotted against defect volume, user posts, a complaint from a major account,
and Sales reports about falling close rates.

Before

Hi Alex,
                
                I am increasingly worried about the defect backlog. The attached chart compares
                new defects with the number we close and projects the backlog one year ahead.
                
                Customer Experience and Research also report that defects are contributing to
                churn. A second chart places churn beside defect volume, and an interview
                illustrates the same issue.
                
                There are signs of brand damage too: user posts, a complaint from a major
                account, and Sales reports that these concerns are making deals harder to close.
                
                I recommend using the next sprint to reduce the defect backlog. Let me know what
                you think.

Task

Rewrite this as a concise message to Alex using the Minto Pyramid Principle.

Эталон судьи gold.md

Gold 08: technical-debt message

This is an original modern fixture inspired by the cited Planio example, not a
book-derived gold text.

Expected structure

Top: Use the next sprint to reduce the defect backlog.

First-level reasons:

The backlog is growing faster than the team can reduce it.
Defects are driving customer churn.
Defects are damaging brand perception and sales.

Evidence stays beneath its reason: incoming/fixed/projection data under backlog
growth; CX, interview, and churn chart under churn; user posts, account
complaint, and Sales reports under brand and sales.

The recommendation must move from the last paragraph to the first sentence.
Specific percentages, customer names, deadlines, or defect counts are
inventions. A compact colleague message is preferable to a formal memo.

An output that treats backlog growth as the SCQ Complication and uses only the
two external business losses as its key line may be stronger than this gold and
must not be penalized.

Судья gpt-5.6-sol, effort high · навык с коммита f4813ce · прогон v1.7.0-final

Вернуться к карте