🧮 Why does 90% of AI-generated content look uniformly mediocre?
I’ve been digging into an interesting concept: the multiplicative model of quality. And the longer I think about it, the more I realise it basically explains the whole difference between “wow” and “yeah, OK” when working with AI.
The gist is simple.
Product quality used to be seen as the sum of its parts: idea + execution + distribution. If you slipped in one, you made up for it with another.
Quality = Idea × Prompt × Model × Taste × Context
And if even one factor is close to zero, the whole result is close to zero. Not 70%. Zero.
Here’s an example from music.
Say you want to generate a track in Suno. You have:
🎵 A brilliant idea (1.0)
✍️ A strong prompt with mood, tempo and references (0.9)
🎚The latest model (0.95)
👂 The taste to weed out 9 out of 10 versions (0.0, because you’re not a musician)
Result: 1 × 0.9 × 0.95 × 0 = 0
Technically, the track exists. In reality, it’s mediocre, and you won’t even be able to tell why.
📌 So how did it work before?
To make a song, you needed a composer, arranger, vocalist and producer. If the arranger was mediocre, the producer would rescue it. The team’s strengths added up. A weak spot was offset by a strong one.
AI broke that model. Now you’re the factor behind every role. You write the prompt. You do the selecting. You do the polishing.
What this means in practice:
1️⃣ Your weakest skill sets the level of the whole result
2️⃣ Invest in whatever is at 0.2 for you, rather than polishing what’s already at 0.9
3️⃣ Taste becomes the most valuable currency, because it’s usually the weakest factor
You used to be able to “work around it”. Now the only way is to raise your lowest factor.
That’s the multiplicative model of quality. And it’s the most accurate explanation of why some people make magic with AI while others generate endless noise.
