Show HN: Gemini has a "Concrete Bias" against minimalist software (Basecamp vs(genrankengine.com)
genrankengine.com
Show HN: Gemini has a "Concrete Bias" against minimalist software (Basecamp vs
https://www.genrankengine.com/blog/concrete-bias-in-llms/
2 コメント
Author here. I ran an experiment comparing retrieval rankings for 'abstract' SaaS (Basecamp) vs 'visual' SaaS (Monday) on Gemini.
It seems the vector distance between 'Visual Nouns' (Gantt, Board) and generic 'Best PM Tool' prompts is shorter than for abstract concepts (Calm, Simple).
Minimalist tools are effectively invisible unless you explicitly prompt for 'text-based' or 'non-visual' workflows.
Has anyone else noticed this penalty for non-feature-heavy tools?
It seems the vector distance between 'Visual Nouns' (Gantt, Board) and generic 'Best PM Tool' prompts is shorter than for abstract concepts (Calm, Simple).
Minimalist tools are effectively invisible unless you explicitly prompt for 'text-based' or 'non-visual' workflows.
Has anyone else noticed this penalty for non-feature-heavy tools?
real title: concrete bias in llms.
Basically its about building site workers trying to come to terms with ai taking over their jobs. The ai brick layer strikes again.
https://www.genrankengine.com/blog/concrete-bias-in-llms/