How Veridonia works

Veridonia decides what's in your feed in an unusual way, and it's easiest to follow if we build it up from scratch. What comes next is the whole idea as a chain of simple questions, starting from the very basics, no prior knowledge needed. Read them in order and by the end you'll see exactly how it works, and why it had to be this way.

— What is a feed?

— A list of stuff (content). Your Facebook, Instagram, and TikTok feeds, your Twitter/X timeline, the YouTube home page, they're all feeds.

— How are feeds formed?

— By letting some stuff in and keeping the rest out. So every feed is just a way of deciding what gets in.

— What kinds of feeds are there?

— Editorial, chronological, and engagement-optimised.

— Are any of them good?

— It depends on how they're used and what for. Each is good for something. But when it comes to just being informed, actually finding out what's worth knowing, none of them are exceptionally good. Editorial feeds can be great, but a fixed few only cover so much, and they're slow to keep up. Chronological feeds just show the newest first, so they can't tell good stuff from bad. Engagement-optimised feeds chase whatever gets the most reaction, some by tracking your behaviour, some by counting likes and votes, but it comes to the same thing: popularity, which favours the entertaining over the useful; and when they run on ads, the advertiser-friendly over the true. And you can't fix any of them by improving it, because in each case the flaw is the way the feed works.

— What would a good feed look like?

— Lots of good stuff, little bad stuff, everything in it posted in good faith, it self-corrects and scales.

— So who decides what's good?

— Nobody can just be handed that job. Not us, not an algorithm, not a fixed few. There's no way to know what a community finds worth seeing except to ask the community itself. So it has to judge for itself.

— How do we build that?

— By finding the people who are good at judging each kind of stuff, and letting them decide what gets in, in a transparent, legible manner.

— How do we find those people?

— We let random members vote on whether a piece of stuff belongs in the feed. Whoever keeps landing in the majority tends to both know the subject and understand what the community wants. Those are the ones we call experts.

— Why random, and not whoever wants to vote?

— Because if people chose what to vote on, anyone with a stake, the author, their friends, a botnet, would pile onto their own stuff. Drawing the voters at random means nobody can aim at a vote they didn't know they'd be in.

— Why would landing in the majority mean anything, and why would anyone vote carefully?

— Because we set it up so that when your votes tend to land in the majority, you get more of a say in what gets into the feed; when they don't, you lose it. So a careless or bad-faith vote isn't free.

— Isn't that just upvotes with extra steps, then?

— No. Say an upvote is meant to ask 'should this go up.' Most people don't even know that's the question, and there's no cost for getting it wrong, so they tap it for 'I agree' or 'it made me laugh' and it collapses into a popularity signal. And the taps just feed a ranking formula that does the real deciding. Ours asks one defined question, there's a real cost for answering it carelessly, and the people's majority is the decision, not fuel for a formula.

— So we just let the experts decide everything?

— It's not that simple, because it breaks three ways: putting every piece to a real vote doesn't scale, the experts harden into a fixed elite, and they drown, voting on every single thing.

— How do we fix that?

— We split the decision into two independent stages. A broad first stage, drawn from the wider community, cheaply filters out the obvious junk. A small second stage, drawn from the proven experts, makes the final call on what's left. Now it scales, the experts only judge the little that survives, and because who gets into the expert stage is always being re-earned, no fixed elite forms.

— And if a call comes out wrong?

— It gets appealed. The appeal goes to a fresh panel, drawn at random like the first, not back to whoever made the call. So a bad decision gets a real second look from different people.

— And how does anyone know it's all fair?

— Because it's all in the open. Every vote, every change to who counts as an expert, and every decision is logged and visible. You can trace exactly why a piece of stuff made it into the feed or didn't. The trust is in the process, not in whoever runs it.

— Why would anyone volunteer to share stuff and vote?

— The same reason people write Wikipedia. Nobody's paid, yet we all rely on it every day. A feed you can actually trust is worth that same effort, and everyone who pitches in gets a better feed out of it.

What you've read here is a distillation of the whole idea. For the overall philosophy, the design goals, and the technical implementation that follows from them, read the whitepaper.