# Writing

> Notes on creative strategy, competitive intelligence and running paid social at scale.

- **URL**: https://marxx.ai/blog

---

- [Static Ad Examples: 16 Image Ads Brands Kept Running (2026)](/blog/static-ad-examples): 16 real static Facebook and Instagram ads, from an 814-day before/after to Notes-app reviews and memes, grouped by format, with what to steal from each.
- [UGC Ad Examples: 14 Creator Ads Brands Kept Running (2026)](/blog/ugc-ad-examples): 14 real UGC ads from Factor, Foxtale, HelloFresh, Starbucks, Timezone and more, grouped by format, with what makes each one believable and what to steal.
- [Best Facebook Ads Examples: 19 Ads Brands Kept Running (2026)](/blog/best-facebook-ads-examples): 19 real Facebook ads from Liquid Death, Duolingo, boAt, Dot & Key, Airbnb and more, grouped by what they do in the first second, with what to steal from each.
- [Best Carousel Ads: 16 Facebook & Instagram Examples Brands Kept Running (2026)](/blog/best-carousel-ads-facebook-instagram-examples): 16 real Facebook and Instagram carousel ads from Canva, IKEA, Airbnb, Wonderla, SUGAR and more, grouped by the job each one does, with what to steal from each.
- [Connect Meta Ads to Claude Code in One Command](/blog/connect-meta-ads-to-claude-code): Add your Meta ad account to Claude Code with one command, then run weekly briefs, competitor teardowns and fatigue checks from the terminal. No API keys.
- [ROAS Didn't Break Your Campaign. It Just Told You Last.](/blog/why-roas-drift-is-a-symptom-and-what-actually-starts-breaking-first): ROAS didn't break your campaign. It just told you last. By the time the ratio drops, the hook has already decayed, frequency has crept up, and CPMs have been quietly inflating for a week and a half. Here's the actual order things break in, and why watching ROAS first means you always find out too late.
- [The World's First AI Model That Understands Ads Deeply](/blog/ads-analyzer-model): Most AI looks at an ad and sees pixels. Ours sees a hook, an offer, a proof structure, and an argument. Here is how we built it and the honest result that killed our best-sounding feature.
- [Deep learning already runs your ad account](/blog/deep-learning-advertising-platform): The deep learning that decides your results lives inside Meta and Google, not inside the tools you buy. Once you accept that, the question changes from "which ad will win" to "what have we never given the model to test".
- [Similarity is not prediction](/blog/similarity-is-not-prediction): We backtested whether creative embeddings can forecast ad performance. Once brand and objective base rates were controlled for, the signal collapsed. Here is what we do with that.