Billionhands · Social ranking product
AI-assisted rankings without giving up authorship
Billionhands turns opinions into rankings people can create, publish and debate. I worked on the product model behind that loop: where AI should reduce effort, where the creator should stay in control, and how community input should sit alongside a personal point of view.
The product problem
A ranking is information, but it is also a point of view
The obvious AI feature was to let someone type a topic and generate a complete ranking. That solved speed, but it created a more important product problem: if the model chooses the candidates and decides the order, whose opinion is being published?
I reframed the challenge around authorship. The system could do work that is expensive or repetitive, but it should not quietly take over the choices that make the ranking personal.
Research
15+ interviews gave us a better set of product questions
We interviewed more than 15 people while shaping the product. Rather than treating research as validation for an AI feature, I used it to understand the jobs inside ranking creation: finding material, expressing preference, and making something worth sharing with other people.
A person may know what they want to rank without having every candidate in mind.
Let the system help discover candidates and reduce blank-page work.
Selection, removal and ordering are not cleanup—they are the opinion itself.
Keep those actions explicit and fully controlled by the creator.
The product needed participation without silently rewriting what the creator published.
Keep creator order and community votes as two visible, separate signals.
The interaction model
Separate discovery, judgement and social validation
The research led to a simple division of responsibility. The system is strongest at searching and proposing material. The creator owns the ranking. The community reacts after publication.
Search across possibilities, interpret broad intent and propose candidates.
Add, reject, remove and reorder until the ranking represents their point of view.
Vote and participate without replacing the creator’s published order.
Candidate discovery
Make AI assistance behave like an editable search layer
The add-participant flow had to work even before AI entered the picture. Direct search, Add, Added and duplicate states establish a predictable base. Assistance then becomes an escalation when the person wants help finding more options.
The important design detail is that model output never bypasses these states. Suggestions still enter the same selection system as manually found candidates.
Published rankings
Community response should add context, not overwrite authorship
The published view keeps the creator, their ordering and community votes visible together. The ranking remains attributable to one person while the audience can express agreement or disagreement item by item.
What I learned
The useful AI feature was a boundary, not a takeover
The initial question was how much of ranking creation AI could automate. The better question became which work actually benefited from automation.
Search and candidate discovery were expensive enough to justify assistance. Selection, ordering and publication were where the product captured human judgement, so those stayed explicit.
That separation made the experience more coherent: AI helps you find what could belong. You decide what does belong.
