Model evals for GTM
Ship or Skip
AI gets me 90% of the way. I'm measuring the last 10%.
Ship = 5 minutes of edit time or less.
The Rubric
Seven areas I grade every output on. A model ships if it passes all seven in under 5 minutes of edits.
1
Accuracy
Facts, numbers, and claims check out
2
Coverage
Hits all the angles the task requires
3
Grounding
Stays in scope without inventing details
4
Reasoning
Logic flows and arguments hold up
5
Alignment
Follows the brief and brand guidelines
6
Style
Reads like a human wrote it
7
Tone
Matches the voice and audience
GTM Tasks I Test
- Blog Post Long-form content for SEO and thought leadership
- Landing Page Copy Conversion-focused page copy with CTAs
- Email Sequence Multi-touch nurture or onboarding emails
- Social Thread LinkedIn or X thread for engagement
- Ad Copy Paid social and search ad variations
- Case Study Customer story with metrics and quotes
- Sales Deck Outline Slide structure and talking points
- Product Description Feature pages and benefit copy
Steal the rubric
Get the full evaluation template I use. Same consent as the newsletter.
Latest Scorecard
Example data. Replace after first real run.
| Task | Claude Fable 5.1 | GPT-6 Astra | Gemini 4 Argon | Grok 4.7 | Coming Soon |
|---|---|---|---|---|---|
| Blog Post | Ship 4m | Ship 5m | Ship 5m | Ship 4m | TBD |
| Landing Page Copy | Ship 3m | Ship 4m | Ship 4m | Ship 3m | TBD |
| Email Sequence | Ship 5m | Ship 4m | Skip 7m | Ship 4m | TBD |
| Social Thread | Ship 2m | Skip 7m | Ship 3m | Ship 2m | TBD |
| Ad Copy | Ship 3m | Ship 3m | Ship 3m | Skip 6m | TBD |
| Case Study | Skip 8m | Ship 5m | Ship 5m | Ship 5m | TBD |
| Sales Deck Outline | Ship 4m | Ship 4m | Skip 8m | Ship 4m | TBD |
| Product Description | Ship 2m | Ship 3m | Ship 3m | Ship 3m | TBD |
| Cost / task | $0.12 | $0.15 | $0.10 | $0.11 | |
| Judge agreement | 94% | 91% | 89% | 92% | |
| Run date | Oct 28, 2026 | Oct 28, 2026 | Oct 28, 2026 | Oct 28, 2026 |
Coming Soon: v2 Assistants
Muse Creative AI assistant
Instinct Personal AI assistant
Personal AI Assistants Custom-trained assistants
Model Reviews
Deep dives on each model's strengths and weaknesses.
- Example entry
Claude Fable 5.1
Ships clean copy with minimal edits. Best for long-form content.