Where models contribute
AI updates news, aggregates feedback and assists comment moderation. Opinion classification and organization are still planned.
Experiment 04 / Demopulse / 2024–26
Follow political leaders, get sourced facts, and have your say.
01 / Hypothesis
I started by testing a direct-democracy MVP. Moving from public figures to news discussion and back taught me that more flexibility can blur a product’s purpose.
02 / Smallest loop
Follow a public figure.
Read a weekly digest of sourced facts.
Submit a rating and comment.
Moderate and aggregate feedback, not reactions to news.
Sourced public context informs individual feedback. A moderation boundary stops some signals before aggregation. The distribution represents contributions, not political worth or a winner.
Sourced context; Public figure; Individual views; Moderation; Distribution.
03 / Division of labour
AI updates news, aggregates feedback and assists comment moderation. Opinion classification and organization are still planned.
Users choose whom to follow and what to say. I set identity, moderation and privacy rules and remain responsible for the product.
04 / Product evidence
Live screens show the feedback interface and sourced news digest, not evidence of adoption.


Reality check / moderation changed the product
By July 2026, Apple had rejected the app eight times. Review requirements led to accounts, blocking and hiding across the app. Moderation improved, but participation became harder and the service held more sensitive political data.
Is this feedback useful enough to justify the identity and moderation requirements?