You’ve felt it. That quiet unease when an AI gives you an answer that sounds confident, measured, almost wise — and you later realize it was confidently wrong. We’ve all learned to live with that gap. We shrug and say, “Well, it’s a tool.” But what if the gap isn’t a bug? What if it’s the entire foundation?
A developer just posted a project on Hacker News called openmog. It’s a social network where the AI has no views of its own. None. Zero. It answers questions using only what people post and comment, and the “truth” of those answers is graded by upvotes and downvotes from the community. Read that again. The AI doesn’t curate. It doesn’t filter. It doesn’t apply some hidden alignment layer to make things palatable. It just takes what the crowd says and treats it as gospel.
Most people scrolling past this will think: quirky experiment, moving on. That’s the wrong reaction. This is a grenade thrown into the foundations of how we think about knowledge in the age of AI.
The moment you strip away algorithmic curation, you don’t find truth waiting underneath. You find a crowd — and crowds are just algorithms made of flesh.
Think about what we’ve been told for the last decade. The problem with social media, the narrative goes, is that algorithms distort reality. They surface outrage. They create echo chambers. They manipulate what we see. The solution, we’re told, is better algorithms, more transparency, more human oversight. But openmog does something radical: it removes the algorithm entirely. The AI has no agenda. It just reflects. And what it reflects is… whatever the majority decides is true.
That should terrify you.
Because here’s what we’ve been avoiding: objectivity was never hiding behind the algorithm. The algorithm was a mask. Remove it, and you don’t get raw truth — you get raw consensus. And consensus has never been the same thing as truth. The crowd believed the Earth was flat for a long time. The crowd believed in phrenology. The crowd retweets misinformation every single day. When you build a system that equates popularity with reality, you haven’t eliminated bias — you’ve just made the bias democratic.
The openmog developer writes, with a kind of innocent precision, that the AI “has no views of its own.” That’s the hook. That’s the thing that makes you lean in. A neutral AI! One that doesn’t lie! But dig into the architecture and the trap becomes visible. The AI treats every post as absolute truth. Every post. The unhinged rant and the careful analysis. The troll and the expert. They all enter the system as equally valid claims. Then the community votes. And the vote decides which “truth” survives.
This is a self-referential loop. The AI’s objectivity depends entirely on subjective collective validation. It’s epistemology eating its own tail.
And yet — and this is the part that should make you uncomfortable — it’s not that different from how we already operate. Most of what you believe to be true on the internet today is just an upvote you didn’t participate in.
When you search Google, the results are ranked by what the crowd linked to. When you check Wikipedia, the articles reflect what editors agreed on. When you ask ChatGPT a question, the model’s answer is a statistical echo of what humans wrote — which means it’s a reflection of consensus, weighted by training data. The difference is that those systems hide the machinery. They dress consensus in the language of authority. openmog rips the dress off.
That’s the twist. openmog looks like a naive experiment, but it’s actually a mirror. It shows us that the line between “the algorithm decided” and “the crowd decided” was always blurrier than we admitted. The algorithm was never separate from the crowd — it was the crowd, compressed, amplified, and hidden behind a brand.
So who should decide what’s true in a networked world? That’s the question openmog forces you to sit with. Not the AI — it has no views. Not the developer — they built it to be neutral. Not the crowd — the crowd is just a different flavor of tyranny, one that punishes dissent with downvotes instead of code.
The uncomfortable answer is that nobody is in charge of truth anymore. And maybe nobody ever was. The only thing that changed is that we’ve stopped pretending.
If you’re building AI products, using them, or just living in a world where they increasingly shape what you see and believe, openmog is worth your attention — not because it’s the future of social platforms, but because it’s the most honest one. It doesn’t promise you truth. It promises you exactly what you already have: whatever everyone else agrees to call it.
And if that doesn’t make you question everything you’ve read today, you haven’t been paying attention.
FAQ
Q: Isn't this just a gimmick? Why would anyone take this seriously?
A: It's a gimmick the way a Rorschach test is a gimmick. The project itself is simple. What it reveals about how we define truth online is not. Every system that ranks, filters, or curates information is doing some version of what openmog does — it just hides the machinery better.
Q: What does this mean for people building AI products?
A: It means you need to stop pretending your model is neutral. Whether you use algorithmic curation or crowd voting, you're making a political choice about what counts as truth. Own it. The worst position is the one openmog exposes: pretending you have no position while the system quietly encodes one.
Q: So you're saying crowds are dumber than algorithms?
A: No — I'm saying they're the same thing wearing different clothes. Algorithms are crystallized crowd behavior. Crowds are liquid algorithms. Neither is a reliable proxy for truth. The contrarian take is that the real problem isn't bias in AI — it's our refusal to admit that 'truth' on social platforms was always a social construct, and AI just made the construct faster.