Creative Diamond
Diverge, converge — the model behind high-volume creative teams.
## 1. Diverge on inputs — ## 2. Expand with machine-generated variants — ## 3. Test fast, not pretty — ## 4. Converge on evidence — ## 5. Repackage the winner across channels —
Why it works▾
Creative Diamond is the 2026 rule: widen with Clay and 11x, then narrow with PostHog and Claude Code until only the highest-converting angle survives.
When to use▾
Use Creative Diamond when you need a repeatable 2026 creative system for paid social, landing pages, and outbound at volume. It fits teams running Meta Advantage+ Shopping, Google Performance Max, and Framer + Relume, where idea generation is cheap but signal is scarce. Use it when one operator can run the loop with Clay, PostHog, and an AI SDR stack instead of a 12-person creative pod.
Steps in detail▾
1. Diverge on inputs
Start with 50 to 200 inputs, not 5. Pull live signals from Clay, Common Room, Keyplay, and Apollo AI, then have an LLM cluster pain points, objections, and category language into 8 to 12 themes. In 2026, a practical example is Clay enrichment feeding Claude 4.5 Sonnet to score accounts by intent, then using those scores to generate 20 unique hooks for Meta Advantage+ creative briefs in one pass.
2. Expand with machine-generated variants
Turn each theme into 10 to 30 variants across copy, angle, proof, and format. Use AdCreative.ai, Pencil, and Arcads for UGC-style prompts, then let Runway or Google Veo 3 produce short video cuts for the top 5 angles. A 2026 example is an 11x agent drafting outbound variants for each ICP segment while a separate n8n workflow pushes the same angle into Framer hero copy and Google Demand Gen assets.
3. Test fast, not pretty
Ship the variants into Meta Advantage+ Sales, Google Performance Max, and Demand Gen with minimal manual approval. The rule is 1 campaign per angle, 1 variable per test, and 72-hour gates on spend. In 2026, a team can use Trigger.dev to launch a batch of 15 ad sets, each with a distinct Clay-derived message, while Northbeam and PostHog track which creative actually drives qualified traffic, not clicks.
4. Converge on evidence
Cut the field to the 2 or 3 winners using PostHog, Northbeam, and June, then have Claude Code analyse session replays, conversion paths, and drop-off patterns. A 2026 example is exporting PostHog events into Claude Code to summarise which headline, proof point, and CTA cluster produces the highest activation rate, then rewriting the Framer page around the winning angle instead of polishing the loser.
5. Repackage the winner across channels
Take the best angle and apply it everywhere: paid, outbound, landing page, and sales follow-up. Use Attio to sync the winning message into lifecycle stages, and let 11x or Lindy send the same positioning to unresponsive leads. In 2026, a single winning Clay-scored pain point can become a Meta ad, a Framer homepage, a PMax asset set, and a personalised 11x sequence within 1 working day.
Pitfalls▾
- 01Creative volume without Clay or Common Room is just noise, not divergence.
- 02If PostHog and Northbeam are absent, teams will optimise for CTR and call it learning.
- 03Running 20 angles at once without 1-variable tests destroys signal quality.
- 04Using human review as the bottleneck kills speed; Claude Code should do the first pass.
- 05If the winning angle is not repurposed into Framer, Attio, and 11x, the loop leaks value.
Get one interactive framework in your inbox each week
Every Monday. No fluff.