The idea
Each lens is a question, and each question is best answered by a particular blend of networks. The model already knows which networks matter for which question, so it reads the second-by-second networks and resolves them into a single 0–100 read per lens, then tracks how that read rises and falls across the content. No language model is guessing here. The lens is a reduction of the predicted brain response, not an opinion about your script.Which networks drive which lens
| Lens | Reads most from | The question it answers |
|---|---|---|
| Attention | Dorsal Attention, Ventral Attention | Is the viewer locked in? |
| Purchase Intent | Limbic (reward) | Will they want it? |
| Manipulation | Limbic, Ventral Attention vs. Frontoparietal | Is it pressuring more than persuading? |
| Emotion | Limbic | How emotionally charged is this moment? |
| Cognitive Effort | Frontoparietal | How much mental work is this taking? |
| Memory | Default Mode | Will this be remembered and absorbed? |
| Surprise | Ventral Attention | Was that unexpected? |
Why this is the “why”
Because every lens traces back to specific networks, you can always ask why a number is what it is. A high Attention read with a low Memory read means the focus networks are firing but the reflection network isn’t, so the content grabs but doesn’t stick. A high Manipulation read means emotional and salience networks are doing the heavy lifting while the reasoning network stays quiet. The lenses are deliberately overlapping cuts of one underlying response. That’s a feature: it lets you see the same moment from seven angles at once.The 7 networks
The raw signal behind every lens.
The 7 lenses
What each lens analyzes for, with examples.