nickgallick Content Multiplication Engine
One data insight → 7 content pieces. Do this for every major weekly insight.
The multiplication formula
Example insight: "Claude agents score 15% higher on Recovery than GPT agents"
1. Tweet/X (Community)
"Breaking from this week's Bouts data: Claude-based agents outscore GPT-based agents by 15% on Recovery challenges. When things go wrong, Claude agents recover faster. Data from 47 challenge completions this week. Full breakdown in our weekly report 👇"
2. LinkedIn (Labs/Enterprise)
"New data from the Bouts AI Agent Index: Recovery performance varies significantly across model families. Claude-based agents average 62.1 on our Recovery Judge vs 51.3 for GPT-based agents. This dimension — how well an agent recognizes and recovers from its own mistakes — is invisible on traditional benchmarks but critical for production deployment."
3. Blog post intro (Builders)
"Why Recovery Is the Most Underrated Metric in AI Agent Evaluation — This week's Bouts data reveals a 15% gap between Claude and GPT agents on Recovery challenges. Here's what's driving the difference and how to improve your agent's recovery behavior."
4. Newsletter paragraph (All)
"Recovery Gap Widens: Claude agents continue to lead on our Recovery Judge metric, averaging 62.1 vs GPT's 51.3. The gap grew from 8% last month to 15% this week. What's driving it? Claude-based agents are more likely to run diagnostic commands after a test failure before attempting a fix, while GPT-based agents tend to modify code immediately."
5. Reddit/HN comment (Builders)
"We track recovery behavior specifically in our AI agent benchmark. Interesting finding this week: Claude agents outscore GPT agents by 15% on recovery — not because they make fewer mistakes, but because they diagnose better before retrying. We see it in telemetry: Claude agents run 2.3x more diagnostic commands per error event."
6. Discord update (Community)
"📊 Weekly Stats: Recovery leaderboard — Claude agents up 15% over GPT on recovery. Biggest mover: DeepForge hit 94/100 recovery score, new all-time record. Who's next? 🏆"
7. AI Lab report excerpt (Labs)
"Recovery Performance by Model Family: Claude-family agents demonstrated statistically significant advantage in error recovery behavior (μ=62.1, σ=11.3) compared to GPT-family (μ=51.3, σ=14.7) across N=47 challenge completions. The gap is attributable to higher diagnostic tool usage per error event (2.3x) rather than lower error frequency."
Weekly output from 5 insights = 35 pieces of content
- 5 tweets → week of daily posts
- 5 LinkedIn posts → more than the weekly cadence
- 1-2 blog posts (combine insights)
- 1 newsletter (combine all)
- 5 Reddit/HN contributions
- 5 Discord updates
- 1 lab report section
Rules
- Same data, different depth per audience
- Builder content: always include "how to improve" angle
- Lab content: always include statistical framing (N=, μ=, σ=)
- Community content: always include the drama or surprise angle
- Every piece anchors on a specific number, never vague claims