How GenAI Evolves: Same Prompt, Different Output
What eight months of model evolution looks like when you run the same creative brief twice.
How GenAI Evolves: Same Prompt, Different Output
What eight months of model evolution looks like when you run the same creative brief twice.

In September 2025, I built an AI art gallery using Google Flow.
Last week I ran the same prompt again.
The output is not the same.
Not because the direction changed. Because the engine underneath changed everything.
Google Flow now runs on Google Omni. ChatGPT moved to Image 2. Eight months later, same interface, different brain.
V1 delivered a version of the intent. V2 delivered the intent.
The models did not just get faster. They got closer to understanding what you actually mean.
Most creatives are still running last year’s prompt logic on this year’s models. The prompt was written for a less capable engine. The model has since outgrown it.
Your old prompt is not the benchmark. Your new output is.
The full breakdown, including the prompt I used and V1 vs V2 side by side, is on STIRMIND.
Read it here: stirmind.com
Frederick Tadeo is a Dubai-based GenAI Creative Lead with 20+ years of digital experience across Leo Burnett, Publicis, and global brands. Through STIRMIND, his creative platform since 2004, he publishes GenAI workflows and AI creative breakdowns. stirmind.com
메타데이터
- post_id
- 70efe1be0b43
- slug
- how-genai-evolves-same-prompt-different-output-70efe1be0b43
- url
- https://medium.com/@dtadeo/how-genai-evolves-same-prompt-different-output-70efe1be0b43
- canonical_url
- https://medium.com/@dtadeo/how-genai-evolves-same-prompt-different-output-70efe1be0b43
- author_url
- https://medium.com/@dtadeo
- status
- ok
- fetched_at
- 2026-06-09 15:37:30