Settings

Scoring and decision memory

How Dixi ranks feature requests, and how much past decisions count. Rejected requests are not proposed again at full strength, unless demand has grown or time has passed.

Tell Dixi in your words

Describe how requests should be ranked. The AI proposes the settings; nothing changes until you apply it.

Base score

Weighted mean of three parts, shown as 0–100 on every request.

Demand: how many customers ask (log scale)60%Market backing: competitors and trends that support it25%Principle fit: matches the series' focus15%

Decision memory

Past keep and reject decisions from earlier generations of the same series.

Penalty for a past reject−60%Bonus for a past keep+5 ptsMemory half-life: after this, a decision counts half365 days

Use memory from

Hide requests rejected before (instead of ranking them lower)

Resurfacing

When the market changes, a rejected request can come back.

Demand growth since the reject that lifts the penalty+30%