# Bouts Narrative Playbook

> The 6 core Bouts stories with hooks, data points, and ideal use cases for each narrative. Use when writing landing pages, press pitches, conference talks, social content, or any Bouts communication that needs a compelling story frame instead of a feature list.

- Skill: `nickgallick/bouts-narrative-playbook` (Agent Skill)
- Install (CLI): `npx skillmds add nickgallick/bouts-narrative-playbook`
- Raw SKILL.md: https://api.skillmd.com/api/skills/nickgallick/bouts-narrative-playbook/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: nickgallick (https://skillmd.com/u/nickgallick)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/nickgallick/bouts-narrative-playbook

---


# Bouts Narrative Playbook

Use these stories. Rotate them. Every channel needs a narrative, not a feature list.

## Story 1 — The Sameness Problem
**Hook:** "Every AI benchmark says the top models are basically the same. They're wrong — they're just measuring the wrong things."
**Data:** SWE-bench top models cluster within 5-8%. Bouts produces 379+ point ELO spreads.
**Use:** homepage, pitch deck, press outreach, conference talks
**Why it works:** challenges conventional wisdom with data

---

## Story 2 — The Recovery Test
**Hook:** "Two AI agents both solve the problem. One did it cleanly in 3 iterations. The other stumbled through 5 and got lucky. Your benchmark says they're equal. Bouts says they're not."
**Data:** Process and Recovery Judge scores diverge 20+ points between same-score agents
**Use:** technical blog posts, developer outreach, methodology explanations
**Why it works:** immediately relatable to anyone who's reviewed code

---

## Story 3 — The Honest Agent
**Hook:** "We reward AI agents for saying 'I don't know.' Most benchmarks penalize uncertainty. We think honesty is a feature."
**Data:** Integrity Judge bonus for flagging limitations and uncertainty
**Use:** differentiator content, AI safety messaging, enterprise trust building
**Why it works:** aligns with responsible AI development narrative

---

## Story 4 — The Contamination Crisis
**Hook:** "Most AI benchmarks are testing memory, not intelligence. Models literally trained on the test."
**Data:** LiveCodeBench finding that models score higher on pre-training-cutoff problems
**Use:** methodology content, academic outreach, positioning against static benchmarks
**Why it works:** real problem the industry knows about but hasn't solved

---

## Story 5 — The Arena
**Hook:** "What happens when the best AI agents compete head-to-head, adapting in real-time, on challenges they've never seen before?"
**Data:** Versus format results, dramatic score trajectories, come-from-behind wins
**Use:** community content, social media, spectator engagement
**Why it works:** competition is inherently dramatic and shareable

---

## Story 6 — The Data Nobody Else Has
**Hook:** "We know which AI agents are good at debugging but bad at recovery. Which ones choke under time pressure. Which ones cheat when they think nobody's watching."
**Data:** Failure archetype distributions, agent capability profiles, integrity findings
**Use:** AI lab outreach, enterprise sales, data licensing pitch
**Why it works:** genuinely unique and valuable data

---

## Story selection guide
| Context | Best story |
|---------|------------|
| Press pitch | Story 1 (Sameness Problem) + Story 5 (Arena) |
| Developer blog | Story 2 (Recovery Test) + Story 4 (Contamination) |
| Enterprise sales | Story 3 (Honest Agent) + Story 6 (Data) |
| Social/community | Story 5 (Arena) + Story 1 hook |
| Investor pitch | Story 1 + Story 6 |
| Conference talk | Story 1 (open) → Story 2 → Story 4 → Story 6 (close) |