Every "which tool is better" poll on the internet has the same problem: the result says more about who brought the most bots and alt accounts than about what people actually use.
So I built Pollreal: 2–4 candidates, 24 hours, and every vote is one verified human.
How a vote is protected
- Phone once, passkey every time. You verify a phone number once (only a one-way hash is stored). After that, every vote is signed with a passkey on your own device — fingerprint, face or screen lock.
- One vote per device, and VPN / data-centre traffic is refused.
- No bandwagon. While a poll is open, only people who already voted see the split. After 24 hours the result is public, with an age and gender breakdown (small groups hidden).
What you see
- Each candidate gets an AI-written card that summarises the maker's official pages, with links to the sources.
- Results are drawn as one bar — like a fighting-game health bar — so you can see how close it was.
- A ranking per category using the Bradley–Terry model (the method behind Chatbot Arena), with 95% ranges: pollreal.com/rank
Open questions
I wrote down what's in place and what I haven't settled — brigades vs. real fans, bandwagon effects, the ranking method: FAIRNESS.md. I'd love for people to poke holes in it.
Closed polls so far include VS Code vs Vim vs Neovim vs Emacs and PostgreSQL vs MySQL vs MongoDB vs SQLite. Results can be read without signing up.
