- Insights
What 500,000 Brand Campaigns Taught Us About Creative That Performs
Patterns that emerge when you study half a million campaigns at once — composition, color, typography, and the briefs that produce the best work.
- Insights
AI Creative vs. AI Slop: It's the Training Data
The visual gap between brand-grade AI creative and the slop flooding the internet has one cause. A look at how training data shapes output character — and why narrow, high-signal training beats vast and generic.
- Insights
The Death of Stock Photography: 18 Months In
Stock photography as a category has been in the slow part of a fast collapse since mid-2023. A working assessment of where the major libraries actually stand at the midpoint of 2025.
- Insights
The Resolution Bar in 2025: What the Tools Finally Reach
Through 2024 and into 2025, the resolution ceiling on AI image tools moved from 2-megapixel novelty output to production-grade 12-16 megapixel work. A working timeline of when each milestone landed and what it unlocked.
- Insights
Black Forest Labs and the FLUX Model Drop
In August 2024, Black Forest Labs released FLUX — a new open-weights image model from former Stable Diffusion researchers. The release reshaped what the open-source side of the market looked like for the rest of the year.
- Insights
Why Generative Video Was Almost-There for Two Years Before Sora
From Runway Gen-1 in early 2023 to Sora's release in February 2024, generative video was a category that kept looking like it was about to break and never quite did. A look at why, and what changed.
- Insights
Stable Diffusion, Midjourney, DALL·E: What Each Was Actually Good For in 2023
A working retrospective on the three dominant image-generation tools as they stood at the end of 2023. What each was actually useful for in brand creative production, and where each fell short.
- Insights
The Pope in a Puffer Jacket: When AI Broke the Cultural Default
In March 2023, an AI-generated image of Pope Francis in a white Balenciaga puffer went viral. The episode taught brand teams more about generative imagery than any case study from the previous year.